Friese S. Qualitative Data Analysis with ATLAS.ti 3ed 2019

166 QUALITATIVE DATA ANALYSIS WITH ATLAS.TI • Select the smart code in Code Manager. Then click on the Edit Smart Code button in the ribbon of the Code Manager. This opens the Query Tool and you can edit your smart code there. Before you can do this, you first need to learn how to build queries in the Query tool (see Skills training 6.5). SKILLS TRAINING 6.3 GETTING TO KNOW THE CODE CO-OCCURRENCE TABLE For this exercise, I would like to explore the following question: RQ1 Do blog respondents who have children define happiness differently from those without children? If so, how do they define it? If you look at the Code Manager, you will see that the ‘def happiness’ category has five subcategories. The codes indicating that someone is a parent or not are: ‘#fam: have children’ and ‘#fam: don’t have children’. The ‘def happiness’ codes occur within each of the blog entries, but there could also be an AND occurrence in case the comment is only about the definition of happiness. Thus, we need to look for code co-occurrences. To find an answer to the research question, one could run single queries in the Query Tool for each ‘def happiness’ subcategory with the two attribute codes (with or without children). A quicker way of gaining an overview that includes all codes is to create a Code Co-occurrence Table. To remind you, for calculating code co-occurrences, ATLAS.ti combines four of the proximity operators and the Boolean operator AND. • From the main menu select Analyze/Co-Oc Table • As row codes, select all ‘def happiness’ codes. If you type ‘def’ into the search field, the list of codes is filtered to the category and you do not have to scroll. You can click on each box one by one or highlight all ‘def happiness’ codes and click on the space bar. This selects all highlighted codes. • In the list for the column codes, select the smart code we created in Skills training 6.2: ‘#fam: don’t have children’ and the code ‘#fam: have children’. You can see from the table that parents provide a larger variety of definitions than those without children, but you must keep in mind that the blog comments contain answers from 75 parents as compared to 20 writers that do not have children. If two codes co-occur, the cells of the table show two numbers (Figure 6.17). The first one is the count of how often the two codes co-occur; the second is the c-coefficient. The c-coefficient indicates the strength of the relation between two codes, like a correlation coefficient in statistics. As with statistical analysis procedures, you need to make sure that your data fulfill the requirements for running a test. Here you analyze the data of two blogs, and

QUERYING THE DATA AND WRITING MEMOS 167 Figure 6.17 Code Co-occurrence Table clearly the case numbers are too low for the c-coefficient to yield a meaningful result. If you import open-ended questions from 300 survey respondents or more, then it makes much more sense to look at this value. Normally, the value of the c-coefficient is somewhere between 0 and 1. But as you are not dealing with standardized quantitative data, it might be that the value is higher than 1. If this is the case, the cell will be orange. This is most often related to a coding error where you have applied a code redundantly. This means, for instance, you have coded a larger segment with Code A and a smaller one within it, also with Code A. You can find such errors using the Redundant Codings Analyzer. It can be accessed from the Tools & Support tab and the Tools tab in the Code Manager ribbon. When the ratio between the row and the column code is larger than five, the field is marked by a yellow circle. For instance, if the column code has 75 quotations, as in the example, and the row code only 11, the ratio is 75/11 = 6,81. To calculate the coefficient, you first multiply the frequencies and then divide the result by the sum of the frequencies minus the product of the two frequencies (see the formula below). If the frequency of one code is very high compared to the other code, the c-coefficient underestimates the strength of the relation. The yellow circle therefore indicates that it might still be worthwhile to look at the data, even though the coefficient is small.

168 QUALITATIVE DATA ANALYSIS WITH ATLAS.TI For anyone interested in the mathematics behind the c-coefficient The coefficient is based on the ‘normalized co-occurrence’ measure as used for quantitative content analysis. In the case of pairwise co-occurrences – that is, co-citation frequency between two and only two terms k1 and k2 – the C-index is given by: C12 – index: n12/(n1 + n2) – n12 where: * c12 = 0 when n12 = 0, i.e. k1 and k2 do not co-occur (terms are mutually exclusive). * c12 > 0 when n12 > 0, i.e. k1 and k2 co-occur (terms are non-mutually exclusive). * c12 = 1 when n12 = n1 = n2, i.e. k1 and k2 co-occur whenever either term occurs. • The c-coefficient is automatically displayed if you create a Code Co-occurrence Table. You can deactivate it in the ribbon. As the Children & Happiness project is so small, it is not meaningful for this project. • You can also switch around the row codes and the column codes. See the option in the ribbon: Rows ==> Columns. • To look at the data behind the numbers, click on any of the cells that show a result. The quotations are displayed in the bottom pane as shown in Figure 6.17. There are two lists of quotations: one for the selected column code and one for the row code. Recall from Skills training 6.1 that ATLAS.ti can only find quotations and not the overlapping parts of the quotation. Thus, when looking for co-occurring codes, ATLAS.ti always finds two quotations, unless it is an AND occurrence where the same quotation is coded with two codes. When you create a Code Co-occurrence Table you likely have a particular idea what you are looking for. Based on this you will know which list of quotations to read. In the example here, the list with the ‘def happiness’ quotations will help us to answer the research question. Exporting the results. If you want to continue to work with the resulting numbers, you can export the table as an Excel file. The quotations of a selected cell can be exported as well. Clustering. If you have coded multiple segments within a larger segment with the same code – for example, all parts of an interview where the interviewee talks about friendship during childhood (the larger segment) – then you may not want to count these as five co-occurrences but as one. This can be achieved by clicking on the clustering button. • Try out the various options so that you become familiar with the tool. • Try out a few more comparisons. For example, instead of the ‘def happiness’ codes, you can compare those with and without children regarding their belief whether children make them happy or not. These are the red codes ‘children: < happiness’, ‘children: = level of happiness’ and ‘children: > happiness’. • Another question you can explore is whether those with one child (#fam: 1 child) or two or more children (#fam: 2 or more children) report different positive and negative effects of parenting.

QUERYING THE DATA AND WRITING MEMOS 169 SKILLS TRAINING 6.4 GETTING TO KNOW THE CODE-DOCUMENT TABLE Using the Code-Document Tables, you can compare the distribution of code frequencies across documents or groups of documents. The sample project contains 24 documents from a survey import. The way ATLAS.ti imports survey data is case-based. This means that ATLAS.ti creates one document per person. If you conducted interviews with one person at a time, the interview transcripts are also one case and you can use the Code-Document Table for case comparisons. Document groups such as gender, region, age groups, occupation, etc. represent cases at the next higher level and can also be analyzed using the Code-Document Table. Let’s create a Code-Document Table now to find an answer to the following question: RQ2 Compare the reasons the male and female survey respondents have given for not having children. • Open the Code-Document Table from the Analyze tab. • In the search field for codes on the top left-hand side, enter NHC. This filters the list of codes and you can easily select all ‘reasons for not having children’ codes. You can apply the same selection technique as has been explained for the Code Co-occurrence Table. • In the bottom-right-hand field for document groups, select ‘gender::male’ and ‘gender::female’. • Switch the rows and columns by selecting Codes as Rows in the ribbon of the Code-Document Table. Figure 6.18 Code-Document Table

170 QUALITATIVE DATA ANALYSIS WITH ATLAS.TI If you click on a cell, you can read the data that is behind the number (Figure 6.18). For instance, you can read what women wrote about self-centeredness as compared to men. Both groups have mentioned it about the same number of times, but without reading the content, you can’t tell whether they mean it in the same way. You can visually support your interpretations by creating charts in Excel. Thus, a quantitative charting of the data terrain is also possible. If documents are of unequal length or the document groups that you are comparing have an unequal number of members, you need to report relative frequencies rather than absolute frequencies. You can activate relative code or row frequencies in the ribbon of the Code-Document Table. Figure 6.19 Comparing code frequencies within document groups You must read this table from top to bottom along the columns. It shows the distribution of subcategories within a selected document or document group. The results of the table could be summarized as follows: as reasons for not having children, women have mostly mentioned ‘self-centeredness’ (58%) and ‘being there for others’ (25%); for men, selfishness was also the main reason (60%), followed by ‘responsibility’ (20%) and ‘not worth the tradeoff’ and ‘state of the world’ (both 10%). Apart from the fact that these data are completely fictitious, this summary shows that percentages can inflate the results quite a bit. There is only one quote for ‘not worth the trade-off’ and ‘state of the world’; in percentage terms, it looks like a lot more. Use this option with care. It is very useful for larger datasets and if you have a higher number of quotations. With n = 4, as is the case with some student projects, it is not methodologically sound to report that 67% of group A as compared to 33% of group B said ...., when n = 3 and 1.

QUERYING THE DATA AND WRITING MEMOS 171 Figure 6.20 Comparing across document groups – relative row frequencies in the Code-Document Table The table shown in Figure 6.20 must be read from left to right. It allows you to compare the application of codes across documents or document groups. ‘Being there for others’ was only mentioned by women as a reason not to have children, and ‘responsibility’ only by men. All other reasons: ‘not worth the trade-off’, ‘state of the world’ and ‘self-centeredness’ were mentioned by both groups equally or almost as often. Depending on how you view the table – documents or codes in the rows or columns – the relative row or column frequencies refer to either the documents or the codes. Since the two groups of documents contain an unequal number of documents, you need to normalize the data to get accurate results. There are 13 documents in the document group ‘gender::female’ and 11 documents in the document group ‘gender::male’. Since

172 QUALITATIVE DATA ANALYSIS WITH ATLAS.TI documents can vary in length, normalization corrects for the number of codings in the documents based on the longest document. If we look at the results after normalization, ‘reasons for nhc: egocentric’ was slightly more frequently mentioned by male respondents than by females. We also need to normalize when comparing documents 3 and 5, which contain the comments on the parenting blog and the comments on the New York Times Magazine article. Document 3 contains 92 comments and document 5 only 47. Thus, if we were to look at frequencies (both absolute and relative) when comparing these two documents, it would not reflect the true difference (see Figure 6.21). Figure 6.21 Relative code frequencies and unequal document length Without normalizing the data, you get the impression that those who comment on the Belkin blog write much more about the positive and negative effects of parenting than those who comment on the New York Times Magazine article: 69% vs. 31% for negative effects and 81% vs. 19% for positive effects. After normalization, the picture changes. There are more comments on negative effects in the New York Times Magazine article (56% vs. 44%) and still fewer comments on positive effects compared to the Belkin blog, but the actual difference is much smaller (40.7% vs. 59.3%). Without normalization, it was 81% compared to 19%. The reference point for normalization is the document with the highest number of quotations of all selected codes. In this example, it is document 3 with a total of 35 quotations. This means that the number of quotations per code for all other documents is multiplied by the ratio of the sum of all quotations of the reference document and the sum of all quotations of the respective other document. In the above example, this is 35/12 = 2,92. • Practice working with the Code-Document Table by examining the comments of the two blogs for other issues, like negative effects of parenting or reasons for or against having children.

QUERYING THE DATA AND WRITING MEMOS 173 SKILLS TRAINING 6.5 CREATING QUERIES IN THE QUERY TOOL The Query Tool lets you formulate code-based queries and retrieves the associated quotations. The Query Tool ribbon contains all available operators for querying the data (see Skills training 6.1) plus a few options that help you in building a query, like adding a code, changing an operator or saving a smart code. • Open the Query Tool by selecting the Analyze tab and from there the first option Query Tool. Figure 6.22 The Query Tool At the left-hand side you see the list of codes and code groups that can be used as arguments in a query. The main space is reserved for displaying the query and the results. Remember the results of a query in the Query Tool are always quotations. Starting simple: building a query using set operators For the next exercise, I would like to explore the following question: RQ3 Find all quotations where people critique the study design or suggest that maybe the wrong questions have been asked. • Select the operator OR in the ribbon. • Select the code ‘study design: asking the wrong questions’ in the list of codes and double-click or drag and drop it into the node with the blue frame around it. The blue frame indicates the currently active node.

174 QUALITATIVE DATA ANALYSIS WITH ATLAS.TI • Click on the second place holder to make it the active node. • Select the code ‘study design: critique’ in the list of codes and double-click or drag it into the node. Above the query you see the full query term. See Figure 6.23. Figure 6.23 Building a query step by step The bottom part of the window shows the list of quotations that results from the query. You can list the quotations of any element in the query. If you click on the code ‘study design: critique’, only the quotations of this code will be displayed in the result list. If you click on the operator, the results of the query are displayed. The currently active part of the query shows a blue frame. The blue status bar at the bottom of the screen shows how many quotations have been retrieved. It is essential to read, review or listen to the contents of the quotations to be able to interpret them. The number of results can serve as an initial point of orientation and, thus, is also meaningful but only together with the corresponding content. If the previews are not sufficient, further options are to create a report or to read the results in context by clicking on a quotation in the results pane. If you have a big screen you may want to work with a floating Query Tool and place it next to the ATLAS.ti main editor to browse through the results. If your screen is smaller, it may be more convenient to dock the Query Tool window and to place it into a second tab group. This way you can view the Query Tool and the documents side by side. See Skills training 2.4: working with docked and floated windows. Creating a report • To create a customized report, click on the Report button. • Select Content and Codes, in case you want to see which other codes are linked to the quotations. • Click Create Report.

QUERYING THE DATA AND WRITING MEMOS 175 Figure 6.24 Creating a report By the time you read this, you may also find an Excel-report option in the Query Tool. As we will need this query for building a more complex query later, we can save it as smart code: • Click on the Smart Code button in the toolbar and enter as code name: ‘study design: raising doubts’. Figure 6.25 Creating a smart code in the Query Tool

176 QUALITATIVE DATA ANALYSIS WITH ATLAS.TI Building a query using proximity operators For practicing the proximity operators, we can explore the following question: RQ4 Find all quotations where people with more than two children describe their own definition of happiness. • Delete the previous query by making the OR operator the active node. Click on the Delete button in the ribbon. • Select the operator Co-occurs in the ribbon. • As we want to read only the segments where those with two children talk about how to define happiness, we first need to select the code group ‘definition happiness’. If we were to enter the code ‘fam: 2 or more children’ on the left-hand side, we would get the entire response of all respondents with two children and not only the parts where they talk about the definition of happiness. See Skills training 6.1. • Click on the second place holder to make it the active node. Select the code ‘#fam: 2 or more children’ in the list of codes and double-click or drag and drop the code into the active node. This query retrieves six quotations. Figure 6.26 Building a code co-occurrence query If you got the codes the wrong way around, you can swap them by clicking on the Swap button in the ribbon. Building more complex queries To find an answer to the next question, we need a combination of two operators: RQ5 Do parents who report negative aspects of parenting also talk about positive aspects? What we want to find are comments written by those who have children (code: ‘#fam: have children’) that contain quotations that have been coded with any of the codes from the category ‘effects pos’ and with any of the codes from the category ‘effects neg’. The codes may overlap or they are embedded within the comment.

QUERYING THE DATA AND WRITING MEMOS 177 Figure 6.27 Illustration of what we want to find with RQ4 We can retrieve all quotations of these two categories via their respective code group. First, let’s find all blog posts where parents write about negative effects of parenting. • Delete the previous query. • Start by selecting the operator: Encloses. • Select the ‘#fam: have children’ code via a double-click to be entered in the node on the left-hand side. The attribute code has always been applied to the entire comment. As we want to retrieve the entire comment, we need to enter the ‘#fam: have children’ code on the left-hand side of the query. • On the right-hand side, enter the code group ‘effects of parenting negative’. See Figure 6.28. Figure 6.28 Query for RQ4 part 1 As we only want to find comments where parents raise both positive and negative issues around parenting, we need to extend the query further: • Make sure that the Encloses node is the active element in the query. You will see a blue frame around it. Now select the Co-occurs operator. • Select the code group ‘effects of parenting positive’ to complete the query. Figure 6.29 Query for RQ4 part 2

178 QUALITATIVE DATA ANALYSIS WITH ATLAS.TI The resulting number of quotations for this query is eight. Thus, eight parents have written about both positive and negative effects of parenting. • Click on the quotations to see the results in context. You will see that all eight quotations include the complete commentary. Now you should be ready to build your own query. Here is your challenge: RQ6 How do people who question the study design see the relationship between happiness and children? To answer this question, you need the following codes: • #blog post • study design: asking the wrong question/study design: critique, or the previously created smart code: study design: raising doubts • children: < happiness • children: = level of happiness • children: > happiness • children: unrelated to personal happiness Hints: you need to use both the Query Tool and the Code Co-occurrence Table, and you need to create a smart code. You will find the solution to the problem at the end of the chapter. SKILLS TRAINING 6.6 LEARNING ABOUT CODE QUERIES IN COMBINATION WITH DOCUMENT ATTRIBUTES For this exercise, we will use the survey data in the project as we have lots of document attributes (= document groups) to play with. Let’s explore the following question: RQ7 Compare the answers of male and female respondents regarding the reasons given for wanting children. So far, we have worked with the blog posts of documents 3 and 5. As these are written by several different people, they needed to be coded for attributes like gender, whether they have children or not, etc. This is also the case when you work with focus group data or if you have documents that contain multiple actors that you want to keep track of. If you look at D7 to D30 in the sample project, you will see that these are case-based data generated through the survey import option (see Skills training 3.7). When working with case-based data like surveys or interview transcripts, there is no need to code for attributes like gender, profession, marital status and the like. This is handled via document groups in ATLAS.ti and always applies if an attribute encompasses the entire document. Thus, if you catch yourself wanting to code the entire document, stop and think of an appropriate document group that you can add this document to. When importing survey data, the document groups are already created, and the documents are sorted into the respective groups based on the information provided in the Excel table.

QUERYING THE DATA AND WRITING MEMOS 179 • Just to get a feeling for the data, open the Document Manager and select, for instance, case 6. It is a single woman with some college education, who has no children and who answered the single-choice question whether children bring happiness or not with ‘Yes’. Figure 6.30 Document attributes for case 6 You are already familiar with the Code-Document Table, which could also be used to answer this research question. An alternative option is to use the Query Tool in combination with the Scope Tool. • Delete the last query. • Double-click on the code group ‘reasons for having children’. • Click on the Scope button and select the document group ‘gender::female’. Read the quotations. The results are always shown in the Query Tool window not the Scope window. See Figure 6.31. • After reading the quotations or creating a report, change the scope to ‘gender::male’. To do that, make sure you are in the Scope Tool. Check the ribbon. Delete the scope ‘gender::female’ and add the scope ‘gender::male’. • Read the quotations of all male respondents to see if they give reasons other than women give, how they present them and what arguments they use. Even if this is just a simple query, by comparing and contrasting responses or statements of people from different groups, lots of insights are often gained. Query results can also be restricted to single documents or a combination of documents and document groups. You may have noticed the set operators in the Scope Tool. In the

180 QUALITATIVE DATA ANALYSIS WITH ATLAS.TI Figure 6.31 Restricting results to a subset in the data next exercise, I would like to show you how to compare the answers of married and single women. For this we need to build a query in the Scope Tool. Before you read on, think for a moment which operator we need to create a new subset for ‘married women’ and another for ‘single women’. You’re right: we need the intersection between all female and all married respondents, which means we need to use the AND operator. • Delete the scope for ‘gender::male’. • Select the AND operator. Add the document groups ‘gender::female’ and ‘marital status::married’ to the nodes. The order does not matter. • The results show four quotations. Figure 6.32 Creating a scope based on two document groups

QUERYING THE DATA AND WRITING MEMOS 181 • Change the scope to single women by deleting the node ‘marital status::married’ and adding the document group ‘martial status::single’. The results also show four quotations. If you need a certain combination of groups more often, you can also create smart groups instead of having to create them anew in the Scope Tool. How this is done is shown in the next skills training. SKILLS TRAINING 6.7 CREATING SMART GROUPS You have already learned about smart codes as a stored query. The logic for smart groups is the same as for smart codes (see Skills training 6.2). If you create smart groups and make changes to your database, the smart groups always reflect the current state of your database. To create a smart group: • Open the Document Manager. • Select two or more groups in the side panel. Check the operator setting that you see on top of the document list. You can switch between ANY (OR) and ALL (AND). • Right-click on one of the selected documents and select New Smart Group. Figure 6.33 Creating smart groups

182 QUALITATIVE DATA ANALYSIS WITH ATLAS.TI A concrete example, as shown in Figure 6.33, would be: • Select the two groups ‘gender::female’ and ‘marital status::single’ by holding down the Ctrl key. • Check that the operator is set to ALL, right-click and select New Smart Group. SKILLS TRAINING 6.8 WORKING WITH GLOBAL FILTERS Global filters are a powerful tool for analyzing your data. Compared to local filters, which you can set in any manager, global filters affect the entire project. If you set a document group as a global filter, the results of the Code-Document Table or Code Co-occurrence Table are also affected. For example, if you cross-tabulate the code category ACTION with the code category OUTCOME, the result refers to all the data in your project. If you want to restrict it to only a subset of your data, you can set a global filter – for example, ‘gender::female’. The results in the table change accordingly and only show the relevant relationships for all female respondents. In a Code-Document Table, you can use global filters to combine two variables without having to create a smart group. Suppose you have created a Code-Document Table for all ACTION codes and the two document groups ‘have children’ and ‘do not have children’. If you set the global filter to ‘gender::female’, the results change in the table and the quotation frequencies only refer to ‘female respondents with children’ and ‘female respondents without children’. An additional benefit is that you can focus your analysis on certain aspects of your project. All selection lists become shorter, and if you know that for the moment you only want to work with five out of your 15 categories, you create a code group that only contains the codes that you want to work with and set this code group as a global filter. Similarly, you can create document groups if you want to focus on only a subset of your data for a given analytic task. Currently, only groups can be set a global filter. In later versions, you will probably also be able to set global filters based on single entities. For now, this means you need to create a group first when you want to use a global filter. Document groups as global filter have an effect on documents and their quotations. Code groups and memo groups as global filter have an effect on code/memos (respectively) only. Enough explaining. It’s best to find out how it works by doing it yourself. Here is another research question to explore. The survey contained two yes/no questions. • Do you believe that children bring fulfillment and give life a purpose? • Do you believe that children bring happiness? Based on the answer, the respondents were added to a respective document group: ‘survey question: bring fulfillment & purpose’ and ‘survey question: bring happiness’.

QUERYING THE DATA AND WRITING MEMOS 183 • Open the document group branch in the Project Explorer and look for those groups. • Open the Quotation Manager. Here is the research question. RQ8 Do respondents differ in their reasons for having children according to whether they mean that children are more about joy or more about fulfillment and purpose in life? • Select all reasons for having children codes (reasons for hc) in the side panel of the Quotation Manager. Enter hc in the search field, so that only the codes for reasons for having/not having children are listed. • In the Project Explorer, right-click on the document group ‘survey question: bring happiness’ and select the option Set Global Filter. • Look at the results in the Quotation Manager and read the quotations. You see which of the codes have been applied in the Codes column next to the quotations or create a report that also contains all codes that have been applied. Figure 6.34 Global filters in the Quotation Manager Your screen should look like Figure 6.34. Since we’ve set a global document filter, you’ll see a blue bar and an orange bar at the top of the Project Explorer, indicating that documents and quotations are being filtered. The codes selected in the side panel of the Quotation Manager filter the list of quotations. As explained in Skills training 3.2, this is a local filter setting as it only affects the Quotation

184 QUALITATIVE DATA ANALYSIS WITH ATLAS.TI Manager. Local filters are indicated by the light-yellow bar above the list of quotations. As the global document filter also affects which quotations are displayed, you see a second bar above it indicating the global filter. The results show all quotations of the selected codes (= local filter) but only for those documents contained in the global filter. • Change the global filter in the Project Explorer to ‘survey question: bring fulfillment & purpose’ by right-clicking on the document group and selecting Set Global Filter. This automatically clears the previous filter. Read the quotations and compare. Although the pictures in this book are printed in black and white, I would still like to point out the color scheme for global filters. These are the entity colors used elsewhere in the software: • Global document filters are blue. • Global quotations filters are orange. • Global code filters are green. • Global memo filters are magenta. • Global network filters are purple. Figure 6.35 Global filter color scheme You can temporarily deactivate a filter by unchecking the box. To clear a filter, you can either click on the X on the right-hand side of the bar (Figure 6.36) or right-click on the group that is set as global filter and select the option Clear Global Filter. Figure 6.36 How to remove a global filter Global filters in the Code Co-occurrence Table RQ9 Compare the difference between those people with and without children regarding their attitudes to whether children bring happiness. Is there a difference between people

QUERYING THE DATA AND WRITING MEMOS 185 commenting on Belkin’s blog and those commenting on the New York Times Magazine article? • Open the Document Manager and create two document groups, one that holds document 3, ‘Belkin’s blog comments’, and one that holds document 5, ‘NYT Magazine comments’. • Open the Code Co-Occurrence Table. • In the Code Co-occurrence Table, select the codes ‘#fam: don’t have children’ and ‘#fam: have children’ as column codes, and the children < / = > happiness codes as row codes. Figure 6.37 Effect of global filter settings in the Code Co-occurrence Table • Set ‘Belkin’s blog comments’ as global filter in the Project Explorer. Inspect the results. • Change the global filter to document 5 (‘NYT Magazine comments’). Figure 6.36 shows the differences. If you deactivate the filter, you see the results for the entire data set. What we can conclude is that this comparison is not very meaningful, as very few people who have commented on the New York Times Magazine article have expressed their attitude. But it only took a few minutes to find out. Sometimes an idea does not lead to anything, and so much the better if you can figure it out quickly.

186 QUALITATIVE DATA ANALYSIS WITH ATLAS.TI Global filters in the Code-Document Table Figure 6.38 shows two Code-Document Tables. Here the number of reasons mentioned for having and for not having children across different educational levels and gender are compared. In addition to the absolute frequencies, the relative code frequencies are shown. To create the tables: • Open the Code-Document Table. • Select the three documents groups that represent the education levels and the two code groups ‘reasons for having children’ and ‘reasons for not having children’. • Switch the rows and columns as needed. • Set the document group ‘gender::male’ as global filter in the Project Explorer. Review the results. • Change the global filter to the document group ‘gender::female’. Review the results. Figure 6.38 Global filters in the Code-Document Table What you can read from the tables is that men with university degrees give more reasons to have children or not to have children. For women, the opposite picture emerges: those with lower degrees indicate more reasons. You are probably somewhat exhausted by now. This is normal when you’re going uphill, but the view from the top is your reward (Figure 6.39). I hope that at least sometimes you felt, ‘Yeah, I am getting there. This is great. I can see things in my data landscape that I did not dream of when I started to explore it.’ Figure 6.39 Rewarding view from the top

QUERYING THE DATA AND WRITING MEMOS 187 It’s time for a little break before the next skills trainings about visualizing relationships and creating networks and hyperlinks. I will tell you a bit about the use of code groups and the use of numbers when analyzing qualitative data with ATLAS.ti. REFLECTIONS ON THE USE OF CODE GROUPS In Chapter 5, I mentioned that it is best not to use code groups as higher-order category codes in a hierarchical sense. I showed how to go about developing subcategories and how to aggregate codes from a lower level to a conceptual level making use of code groups as a filter tool. In this chapter, you have seen how useful code groups are in formulating queries. For this reason, I generally create code groups from all main categories and their subcategories in the process of developing a code scheme. I advise against using code groups solely for creating categories. This results in an unstructured list of codes that is difficult for you to handle and for others to comprehend. What also often happens when code groups are used to aggregate codes is that early descriptive codes are not developed further into ‘proper’ codes. They are just grouped into a group without being further conceptualized. As I mentioned in Chapter 5, the software does not recognize the different levels of codes. It is up to the software user to make this distinction. If a code remains at the descriptive level and is not conceptualized, there is no warning sound or flashing red light indicating: ‘Please look at this code; it is not yet a “proper” code. You need to work on it a bit more!’ To the software, a code is just an entity that can be attached to various other entities and whose content can be searched and retrieved. Everything else is up to you. A further aspect that needs to be considered is that although code groups can be included in networks, they can be linked neither to each other nor to other codes using named relations (Figure 6.40). Figure 6.40 Code groups in networks You might say that this is not much of a disadvantage as most of the interesting linkages are not on the category but on the subcategory level. I agree. The greater danger I see when

188 QUALITATIVE DATA ANALYSIS WITH ATLAS.TI using code groups as categories is that descriptive label codes are not conceptualized properly and remain unsorted in the Code Manager. This is illustrated in Figure 6.41. A second reason for not using code groups as categories is methodological rigor. The first step should always be to develop the coding scheme in a proper manner. This means that codes should not be all over the place on different levels with no indication of which code belongs where and what kind of meaning it has. If you use the code group function in ATLAS.ti too early, it undermines this process and the groups will get in the way of building a systematic, transparent and comprehensible coding system. Figure 6.41 Display of codes, if you were to use code groups as categories REFLECTIONS ON THE USE OF NUMBERS AND HOW PERFECT DOES THE CODE SYSTEM NEED TO BE? You’ve seen throughout the chapter that quotation frequencies, the number of codes and subcategories provide an initial overview of the content of the data. However, these must be carefully interpreted in qualitative data analysis. It is always good advice to review the data behind the numbers. In addition, remember that the output is only as good as your coding. While preparing the exercises for the book – running through them, testing and reviewing the data – I came across quotations where I wondered why I had coded them the way I did. When the assignment of a code no longer appeared to be meaningful, I unlinked the code, assigned others or modified the segment length. I am sure that this will also happen when you work on your own data. Querying data means you view them from different perspectives; you begin to understand more and your views change on how segments should be coded. Thus, coding and recoding never really finish. If your coding system is built up well,

QUERYING THE DATA AND WRITING MEMOS 189 the changes are not huge; however, subtle changes are likely. The hermeneutic circle continues. The more you work with your data, the better you understand what is going on and the closer you approach the underlying meaning. Thus, when you stop your main coding phase to begin querying the data, keep in mind that your coding system does not have to be perfect: 95% perfect is good enough – don’t spend another three months trying to reach 99% perfection. As the coding system in most studies is something that continuously changes with progressive analysis, this tells you something about how to make use of and interpret the numbers that ATLAS.ti provides. Use them as indicators but not as facts. What I hope you have seen in this chapter is that only a well-built-up code system based on categories with subcategories allows you to ask the kind of queries that you have gone through in this chapter. You learned how to click on queries in the Query Tool; how to create cross-tabulations of codes and cross-tabulations of codes by documents and document groups; how to combine code queries with variables; how to create special filters in the form of groups and smart groups; and how to save queries for later reuse in the form of smart codes. You also saw that certain types of data require the use of different tools depending on how you need to deal with data attributes (creating document groups or coding them). Now that you know which analysis tools are available for retrieving data, it’s probably easier to see why in Chapter 5 I give specific recommendations on how to build an efficient coding system. REVIEW QUESTIONS Here are the questions for you to work through: 1 Which operators are available to create queries? Explain them. 2 How do you store a query for later reuse? 3 Which function do you need to create a frequency table showing code frequencies by documents or document groups? 4 When do you need relative row or column frequencies? 5 What type of analysis can you run using the Code Co-occurrence Table? 6 What does the c-coefficient indicate? When is it useful to use it? 7 How do you build a query in the Query Tool? 8 How can code searches be combined with variables? 9 How do you create and apply filters to restrict a query to a subset of your data? 10 How do you create reports from query results? 11 What is the purpose of global filters and how do you build a query with global filters? 12 Why do you need to use different analysis tools depending on the type of data you have? Fortunately, if you cannot answer everything, this is a book and you can go through the chapter again at your own pace and repeat where you are still unsure. Use the provided sample data and click on the instructions again to see what happens on the screen. Or wait until you have encoded your own data to apply the analysis tools to your own questions. Often

190 QUALITATIVE DATA ANALYSIS WITH ATLAS.TI the ‘aha’ moment comes when you use the analysis tools on your own data and suddenly everything becomes clear. SOLUTIONS Skills training 6.1: understanding Boolean operators Figure 6.42 Solutions to Boolean queries Skills training 6.5: step-by-step instruction to answer RQ6 RQ6 How do people who question the study design see the relationship between happiness and children? • You can either build the first part for this query from scratch or use the smart code that we created earlier in Skills training 6.5: Figure 6.43 Solution to RQ5 part 1

QUERYING THE DATA AND WRITING MEMOS 191 • This query leads to 23 quotes. Save the query as smart code – for example, under the name ‘*respondents who question the study design’. I’ve added an asterisk (*) at the beginning of the code label so that the code is listed in the upper part of the code list and I can easily find it. It is ‘not lost’ somewhere between the thematic codes. • Open the Code Co-occurrence Table and create a table with the smart code as column code and the four attitude codes as row codes: Figure 6.44 Last step to answering RQ6 – creating a Code Co-occurrence Table You may be a little disappointed with the results after trying so hard to create the query. Remember, it was just an exercise :-). What we can read from the table is that most of those who question the study design did not mention how they see the relationship between happiness and children. When they did, they expressed a neutral attitude. FURTHER READING Ayres, Lioness, Kavanaugh, Karen and Knafl, Kathleen (2003). Within-case and across-case approaches to qualitative data analysis. Qualitative Health Research, 13(6), 871–83. Bazeley, Pat (2002). Issues in mixing qualitative and quantitative approaches to research. Presented at the 1st International Conference on Qualitative Research in Marketing and Management, University of Economics and Business Administration, Vienna, www. researchsupport.com.au/MMIssues.pdf. Freeman, Melissa (2017). Modes of Thinking for Qualitative Data Analysis. New York: Routledge. Kelle, Udo (ed.) (1995). Computer-aided Qualitative Data Analysis, part iii. London: Sage. Miles, Matthew B., Huberman, A. Michael and Saldaña, Johnny (2014). Qualitative Data Analysis, 3rd edn, chapters 6–11. Thousand Oaks, CA: Sage. Neri de Souza, Francislê, Neri, Dayse Cristine and Costa, António Pedro (2016). Asking questions in the qualitative research context. The Qualitative Report, 21(13), 6–18, http:// nsuworks.nova.edu/tqr/vol21/iss13/2. Richards, Lyn (2009). Handling Qualitative Data: A Practical Guide, chapters 7–9. London: Sage. Silver, Christina and Lewins, Ann (2014). Using Software in Qualitative Research: A Step-by-step Guide, 2nd edn, chapter 13. London: Sage.

Recognizing and visualizing relationships – working with networks 7 The ATLAS.ti network function is a tool that allows you to explore your data visually. You started the journey by looking at an unknown landscape (see Chapter 3). Then you began to explore it by noticing interesting things. After a while, you were able to label what you noticed, and you gained a better understanding of the data landscape during the process of coding through Chapter 5. At first, this was a descriptive understanding. With a prolonged stay and further exploration, you were able to describe the various aspects of our data landscape and their specifications in the form of a well-developed code system. This enabled you to dig a bit deeper and to ask more specific questions utilizing several different tools provided by the ATLAS.ti workbench, as explained in Chapter 6. As you go through your research questions and queries step by step in the process of further analysis, writing memos is a must. By writing, you need to put into words the results you see in the form of tables, numbers and data, and add your thoughts, ideas and interpretations. In this process analysis ‘happens’. As Freeman (2017: 4) puts it: it is important to understand that ‘writing is inseparable from analysis’. It is also during this process that you realize how the different aspects of your data relate to each other. The visualization of these relationships in the form of the ATLAS.ti networks is a next logical step. It will advance your analysis on another level. Graphical illustrations

RECOGNIZING AND VISUALIZING RELATIONSHIPS – WORKING WITH NETWORKS 193 enable a different kind of exploration. Images activate different parts of the brain from words and lead to different processing modes (see, e.g. Khateb et al., 2002). The ATLAS.ti networks support creativity and help in the detailing of an idea or by developing a line of reasoning. They improve metacognition by encouraging a different way of thinking. At the receiving end, they help create a common understanding and help communicate complex ideas and arguments (Freeman, 2017; Novak and Cañas, 2006; Novak and Gowin, 2002). Figure 7.1 shows the story that can be told about our imaginary data landscape. You can see which people belong together, who is living where and related to whom, and the secret attraction of this valley: the cairn of wisdom. Not everybody has access to the cairn of wisdom; some can only observe from the outside while others undergo various rituals to gain entry – though this requires such mental strength and willpower that not all inhabitants of the valley or its visitors will succeed… If you enjoy storytelling, look at the concept map laid over the data landscape and try to continue the story. You can embellish it with further detail, mystery and possibly a happy ending. Transferring this to your own project, as a result of your data analysis, you can tell the stories contained in your data through ATLAS.ti networks. Figure 7.1 Concept map of the data landscape In this chapter, I will first provide some ideas about the purposes for which the network function can be used. After this introduction, I will explain the technicalities of working with networks. Once again, this will require some skills training. You will need to learn

194 QUALITATIVE DATA ANALYSIS WITH ATLAS.TI some new terminology and on the technical side how to create links and relations. In this chapter, the order in which you go through the skills training after Skills training 7.1 doesn’t matter. Some may prefer to start with the technicalities (Skills training 7.3); others may like the pragmatic approach and want to learn about the application of networks first. LEARNING OBJECTIVES In this chapter, you will learn a few more technical skills: how to create networks, how to link the various entities, how to create your own relationships and how to export networks so that you can use them in reports. You will also learn how networks can be used for analytical purposes and how visualizing relationships can help answer research questions. In the last part of the chapter, I’ll show you how to use and create hyperlinks. SKILLS TRAININGS Skills training 7.1: learning terminology Skills training 7.2: using networks for conceptual-level analysis Skills training 7.3: creating and working with networks Skills training 7.4: working with the Relations Editor Skills training 7.5: working with hyperlinks SKILLS TRAINING 7.1 LEARNING TERMINOLOGY Basically, everything you do in ATLAS.ti – each link you create, be it a code–quotation link, a code–memo link, a memo–quotation link, a code group linked to its code members and so on – can be visualized. The entire project consists of links and, thus, represents the total network. The process of coding has already generated a great number of links between individual codes and the data segments they encode. Figure 7.2 shows the code ‘effects pos: improved relationships’, which codes six data segments and is linked to two other codes. Incidentally, each object that becomes part of a network is called a node. As you can see from the network here, there are different types of relations – just a line linking two objects or a line plus a name for the link. The named links are referred to as first-class relations and the unnamed links as second-class relations. First-class relations can only be created between two codes and between two quotations. All other links are second class.

RECOGNIZING AND VISUALIZING RELATIONSHIPS – WORKING WITH NETWORKS 195 Figure 7.2 First- and second-class relations First-class relations can be directed – that is, point from A to B (A -> B) – or undirected (A <--> B). In Figure 7.3, you can see an example of each one. The property of directed relations is transitive or asymmetric and those of undirected relations symmetric. You need to know these terms when creating new relations, as I will show in Skills training 7.5. Figure 7.3 Examples of the various possible links and visualization options

196 QUALITATIVE DATA ANALYSIS WITH ATLAS.TI Second-class relations can be created between memos and quotations, codes and quotations and memos and other memos. The links between groups and their members can be visualized as well. Indirect connections exist between codes and documents and codes and document groups (see Figure 7.12). All these links are second-class – thus, they cannot be named. When you open a network on a group, the lines between the group node and the member nodes are shown automatically. The same happens if you import quotations or codes for a document. In Figure 7.3, you can see an image document that has been added to the network and the connection to a code (the dotted line). You can choose from several View options. Figure 7.3 only shows quotation IDs and names, but you can extend the display up to the full text for each quotation. Code nodes can be shown with and without their groundedness and density counts (Figures 7.7 and 7.8). If you don’t like the node icons, you can deactivate them. You will explore the available options in Skills training 7.3. In addition to the various layout options, several actions are also available. Networks do not simply display your project items: you can also access the data behind them. Doubleclicking on a memo in a network will display the memo text. When double-clicking on a quotation, you can view it in the document. When double-clicking on a code, you have the option to list all quotations coded by that code and to access the data from the list (Figure 7.23). SKILLS TRAINING 7.2 USING NETWORKS FOR CONCEPTUAL-LEVEL ANALYSIS I would like to introduce this functionality to you in a pragmatic manner by going through a few examples without first telling you the more boring technicalities of the network function. This way, you will learn some of the mechanical stuff just by doing it. If you are a more theoretical kind of person, you may want to go through Skills trainings 7.4 and 7.5 first and then return to this section. Exploring code co-occurrences in networks If you want to click along, you can continue to work with the Children & Happiness_ analysis (chapter 6 to 8) project from the companion website. It is a coded project that does not yet contain links. • Open the Code Manager. Select the code ‘#fam: 2 or more children’, right-click and select Open Network from the context menu. A network opens that contains only one code. • Drag and drop the code into the top-left corner of the network. Right-click and select Add co-occurring Codes. Since we have not set a filter, all codes that co-occur at least once with the code ‘#fam: 2 or more children’ are added to the network. With so many codes in the network, it’s unlikely that we will see anything meaningful.

RECOGNIZING AND VISUALIZING RELATIONSHIPS – WORKING WITH NETWORKS 197 To produce the image you see in Figure 7.4 the perpendicular layout was chosen (see Layout button in the ribbon). Figure 7.4 Adding co-occurring codes without setting a filter Figure 7.5 Global code group filter for Skills training 7.2 To create a more meaningful network: • Create a new code group with the name ‘*skills training 7.2’. Add the following codes to it: ‘#fam: 1 child’, ‘#fam: 2 or more children’ and all subcategory codes of the category EFFECTS POS (see Figure 7.5). • Set this new code group as global filter (right-click on the code group and select the option Set Global Filter):

198 QUALITATIVE DATA ANALYSIS WITH ATLAS.TI • Right-click on the code ‘#fam: 1 child’ and select Open Network. • In the network, right-click on the code node and select Add Co-occuring Codes. Now only three codes are added to the network. • Repeat this for the code ‘#fam: 2 or more children’. Five codes are added to the network. If you place the two networks side by side, you can make a meaningful comparison. Figure 7.6 Adding co-occuring codes using a global filter setting – comparing different groups of respondents regarding the positive effects of parenting they wrote about Let’s look at RQ1 again: do blog respondents who have children define happiness differently from those without children? We have already explored the question in Chapter 6. I will show you how we can find an answer to it using the network function and the Code Co-occurrence Explorer, which we haven’t used so far. First, we need to create a code group that contains the codes we need for this research question. • Create a code group with the name ‘*RQ1’ and add the codes ‘#fam: have children’, the smart code ‘#fam: don’t have children’ and all subcategory codes of the category ‘DEFINITION HAPPINESS’ (see Figure 7.7). • Set this code group as a global filter. • Open the Code Co-occurrence Explorer in the navigator. To do so, click on the drop-down menu of the Navigator button in the Home ribbon. • Open the branches for the two codes ‘#fam: don’t have children’ and ‘#fam: have children’ (see Figure 7.8). Now ATLAS.ti runs the same query as it does when preparing a Code Co-occurrence Table or when you use the COOCCUR operator in the Query Tool. Figure 7.7 Code group *RQ1 set as global filter

RECOGNIZING AND VISUALIZING RELATIONSHIPS – WORKING WITH NETWORKS 199 You immediately see the difference between those who have children and those who don’t. Figure 7.8 Opening and expanding the Code Co-occurrence Explorer • To represent this in a network, you can link the green definition happiness codes to the gray ‘#fam: don’t have children’ and ‘#fam: have children’ codes via drag and drop (see Figure 7.9). Select the relation ‘is associated with’. Figure 7.9 Linking codes in the Code Co-occurrence Explorer

200 QUALITATIVE DATA ANALYSIS WITH ATLAS.TI • Next, highlight all codes, right-click and select Open Network. • Arrange the codes in the best way to see the difference between those who have and do not have children. • For the final layout, click on the Routing button and select Polyline Routing. Figure 7.10 shows a possible layout for visualizing the findings of this research question. Figure 7.10 Visualizing of results of a research question In Figure 7.10, you can see the frequency and density counts below the code labels. The density count shows how often a code is linked to other codes. To activate this option, select the View tab and from there Show Frequencies. Whether the network already tells the complete story can only be discovered after reading the data behind the code nodes. This can best be achieved in combination with the Code Co-occurrence Table. See Skills training 6.3 and Figure 6.17. After reading the data behind the nodes in the network and finding that the links are correct and that the relations apply, you can save the network. This way, you preserve the view in the way you have created it and can always review it later or continue to work on it. • To save the network, click on the Save button in the ribbon and enter a name for the view like ‘RQ1: Differences in defining happiness by respondents with and without children’. If you close the network without saving it, ATLAS.ti will remind you to save it. (How to export a network as a graphic file to insert it into a report or presentation is explained in Skills training 8.3.) Case-based analysis in networks Another option is the entity sensitive import. Depending on which entity you select (code, quotation, memo or document), the Add Neighbors submenu shows different options.

RECOGNIZING AND VISUALIZING RELATIONSHIPS – WORKING WITH NETWORKS 201 When combining this import option with global filters, it is, for example, possible to compare documents in terms of their coding. Let’s take a new look at RQ3 and explore it in a network: RQ3: Compare the comments written on Belkin’s blog to those written on the New York Times Magazine blog regarding, for instance, sources of happiness. • Set the ‘Sources of happiness’ code group as the global filter in the Project Explorer. • Hold down the Ctrl key and select D3 (Belkin’s blog) and D5 (NYT Magazine discussion) in the Project Explorer. Right-click and select Open Network. • A network editor will open displaying the two documents. If you want to see a preview of the documents, click on the View tab and select Preview. • To see which of the ‘sources’ codes have been used in document 3, right- click on D3 and select the option Add Neighbors / codes. Repeat this for D5. Codes that have been applied to the documents are linked with a dotted blue line. If you don’t see it, check in the View tab whether the option code-document connections is activated. • Position the nodes either manually or use one of the layout options. Arrange the two documents so you can see whether there is a difference between them. In the View tab, you will find the option Snap, which helps you to align the nodes in a network. Figure 7.11 Case-based network Let’s extend the research question a bit and add document 6 to the network. D6 contains summaries of research findings regarding sources of happiness. By adding this document

202 QUALITATIVE DATA ANALYSIS WITH ATLAS.TI you can see if the people who wrote comments on the blogs differ in content from scientific research findings: • To add D6 to the network, drag and drop it from the Project Explorer into the network. If you have a small screen, you may want to deactivate the preview. • To see whether it contains further codes, you also need to import the codes for this document. Thus, right-click D6 in the network and select Add Neighbors / codes once again. • Arrange the network in such a way that you can see which of the ‘sources’ codes apply to all three documents, which to only two documents and which to only one of the three. The network shows that most of the scientifically proven sources that contribute to a sense of personal happiness have also been mentioned by the people commenting on the blogs. What they did not consider is that financial security and just a smile on the face help make you feel happy. Explicitly, the blog writers mentioned the following: children’s contribution to happiness, that happiness must be achieved through struggle and that it is more about enjoying the journey than considering happiness as an end by itself. Using networks to develop the storyline for your research report When I developed the grounded theory sample project based on the Strauss/Corbin approach, I came to the point where I had to choose a core category. The study was about war experiences of veterans. I decided to build the analysis around the core category ‘coming home’. At the time, I already had some ideas on how the ‘coming home’ code could be linked to other concepts and categories. So, I created a network and pulled in my ‘Coming Home’ code. Based on a number of questions I asked myself, I added more codes to the network. What factors make a homecomer feel that he or she has arrived home? What hinders a successful return? Which strategies work against this? Are there differences between those who fought at the front and those who served behind the lines, like the paramedics? Which coping strategies were used during and after the war? What was the original attitude towards the war? Has this attitude changed? In other words, I worked backward in time from the present to the past to develop my storyline. Figure 7.12 shows the network I created. It looks a bit messy. But the aim was not to create a network that can be used in a presentation or report. It just served me and helped me think about possible connections by placing the nodes in the network, creating links and naming relations. Based on this large network, I focused on partial aspects of the story and created several smaller (less chaotic) networks (see Friese, 2016, 2019 for further details). At the same time, I wrote down my thoughts, ideas and interpretations in research-question memos (see Skills training 5.7). Writing memos is also an integral part of this analytical phase. Creating networks helps you see things in the data, but you must write them down, otherwise the ideas that come to your mind will be fleeting and forgotten tomorrow. And as you write, the connections become clearer to you, and gradually the individual parts can be linked and result in a coherent story.

RECOGNIZING AND VISUALIZING RELATIONSHIPS – WORKING WITH NETWORKS 203 Figure 7.12 Messy network to develop a storyline Using networks to discuss findings with your adviser or colleague(s) As already mentioned above, visual images activate different regions in our brain and stimulate us to think in different ways. You may send your adviser or colleagues an excerpt from your analysis chapter before scheduling a meeting, so they can get an idea about your work. Prepare one or two networks that visualize the main arguments that you want to discuss and bring a printout and possibly also your laptop to the next meeting (see Skills training 8.3 on how to prepare a printout). Instead of going through pages of text, you can explain your ideas while looking at and discussing the network(s). If there are questions related to the underlying data, you can open the network(s) in ATLAS.ti on your laptop and access the data from within the network (see Skills training 7.3). Talking about the findings based on the printed-out network might, however, already be enough. Working with a different medium, paper in this case, can be a nice change, especially when it comes to visualizing ideas. You can extend your ideas by scribbling notes on the printout or by drawing new ‘networks’ on paper. After the meeting, you transfer your ideas, notes and paper-based networks to ATLAS.ti. As the data in ATLAS.ti are only a few mouse clicks away, you can verify whether the ideas are valid and still hold when checking them against the data. If so, you can refine your current networks and the analysis you wrote in your research-question memos. Using networks to present findings Below, I present some example networks from various studies to show how networks can be used to present findings. For smaller projects, like a Master’s thesis, it might be possible to integrate all findings into just one network, but sometimes several are needed.

204 QUALITATIVE DATA ANALYSIS WITH ATLAS.TI Illustrating results from the Schwarzenegger project The network shown in Figure 7.13 illustrates a result from analyzing the sample data set used for the first edition of this book. The data consisted of newspaper articles from Germany and the USA collected the day after Arnold Schwarzenegger was elected Governor of California in the 2003 recall election. One can insert networks as a graphic into a PowerPoint or Prezi presentation. Prezi has the advantage of being able to zoom to the parts of the network you are currently talking about. The text that goes along with the network shown here could be something like this. The network shows that there are differences in topics covered by the German local press and the German national press. As can be seen from the network, the local press provided some information on the recall process and strongly focused on the election results. In comparison, the national press provided more general background information and focused more on Schwarzenegger’s political program. Figure 7.13 Network of the issues covered by the German local and national press Illustrating results from a media analysis of the financial crisis Figure 7.14 shows the results of a study comparing media reports from 2008 and 2009 on the financial crisis and illustrates how you can make your data come alive in presentations. This requires you to run ATLAS.ti in the background or, alternatively, to use ATLAS.ti to present your findings. The network below shows factors that have been mentioned by various sources (personal experiences, statistical figures, news agencies and political opinion)

RECOGNIZING AND VISUALIZING RELATIONSHIPS – WORKING WITH NETWORKS 205 as immediate, long-term and individual consequences.1 Also shown is an activated audio quotation that can be played. Text quotations can be viewed as a preview or in the context of the data. Video quotations can also be played. Figure 7.14 Integrating findings from a study on the financial crisis (Friese, 2011) Using networks in publications The following two networks are included in genuine publications and represent central findings. Figure 7.15 shows one result of my dissertation research illustrating the phases of an addictive buying experience. The original network published in my dissertation (Friese, 2000) looked a bit different as several options were not yet available in version 4 of ATLAS.ti. And, as mentioned, I would not have coded my data in the same way as I did back then and, therefore, I have modified the original code labels to reflect how I would probably code the data today. For the next example I am indebted to Eddie Hartmann for allowing me to use his data material (Hartmann, 2011). He conducted 20 interviews and developed a case structure for each person based on four main criteria: negation, affirmation, rejective negation and positive substitution. Figure 7.16 shows one of the cases. One can see overall which criteria applied and within each criteria which subcategories were relevant. ‘Affirmation’, for instance, did not apply to case 3 at all. Quotations were added to each network to show the sequence of arguments that made up the story of each interviewee. 1The study was developed as a sample study based on a small data set. Therefore, the results are fictitious.

206 QUALITATIVE DATA ANALYSIS WITH ATLAS.TI Figure 7.15 Illustrating the addictive buying experience (adapted from Friese, 2000) Figure 7.16 Using the network function to show case patterns in the data SKILLS TRAINING 7.3 CREATING AND WORKING WITH NETWORKS In the next few pages, you will learn about various ways of linking and handling networks. The exercises are intended to teach you the technicalities of working with networks. You will not create ‘meaningful’ networks in an analytic sense. You can choose any of the

RECOGNIZING AND VISUALIZING RELATIONSHIPS – WORKING WITH NETWORKS 207 existing codes to practice with, like you did in Chapter 4 with the four test codes when I showed you how to code. Learning how to link • Open the Code Manager and choose four codes by holding down the Ctrl key. Then click on the network button in the contextual code ribbon or right-click select Open Network from the context menu. • A network editor will open. As a title, the name of the selected code(s) is chosen. The window contains the code you have selected. If you want to add more codes, move the network window next to the Code Manager and drag and drop other codes into the network, or drag and drop codes from the Project Explorer, code list or the margin area. Alternatively, you can click on the Nodes tab and from there select the ribbon button Add Nodes (see Figure 7.22). • To link codes, select one of the code nodes with a left mouse click. A red dot will be displayed in the top left corner. Drag the red dot with your mouse to another code and release the mouse button. A list of relations will open. Select the is associated with relation. Figure 7.17 List of code–code relations Relation properties This relation is a symmetric relation displayed as a line with an arrow at each end. It indicates that the two codes are related but no direction can be specified. Depending on the type of project you are working with, the list of relations on your computer screen might be shorter than the one shown in Figure 7.17. In addition to the default relations that ATLAS.ti provides, it includes a few others that I have created for the sample

208 QUALITATIVE DATA ANALYSIS WITH ATLAS.TI project. Basically, you can create any relation you want and need and in any language (see Skills training 7.4). Let’s link the other codes: • Link two codes using the relation ‘is a property of’. It is a directed asymmetric relation and displays an arrow at the node it is pointing to. • Create one more link using the relation ‘is part of’, which is a directed transitive relation. Figure 7.18 Relation properties The difference between directed and non-directed relations is easy to understand. What needs further explanation is the difference between the two directed relations: asymmetric and transitive. The following equations are transitive: 2 < 4 < 6, then 2 < 6. Another example is: if Tom ‘is ancestor of’ Mary, and Mary ‘is ancestor of’ Paul, then we can also state that Tom ‘is ancestor’ of Paul. If we use the relation Tom ‘is father of’ Mary, and Mary ‘is mother of’ Paul, then it is not true that Tom ‘is father of’ Paul. ‘Is ancestor of’ is a transitive relation, whereas ‘is father of’ is asymmetric. Transitive and asymmetric relationships can be graphically represented as follows: Figure 7.19 Difference between transitive and asymmetric relations

RECOGNIZING AND VISUALIZING RELATIONSHIPS – WORKING WITH NETWORKS 209 Most directed relations in qualitative data analysis will be asymmetric rather than transitive. We may, for example, want to express that if there is context X, then A causes B and that has a consequence C. Not only did we choose a different relation on each link, it is also unlikely that context X will result in C as consequence. An example of a transitive linkage would be: if A causes B and B causes C, it can also be said that A causes C. If this the case, then semantic operators can be used in the Query Tool to explore these linkages. See Skills training 6.1. Exploring the links • The link labels are interactive. Right-click on a link label and explore the options like flipping a link, changing the relation or cutting a link (Figure 7.20). • Each link can be commented individually. Try it. All commented links are marked with a tilde (~) as you already know from other commented entities in ATLAS.ti. • End this exercise by unlinking all codes from each other because I want to show you next how you can link multiple nodes to each other. For this you need free unlinked nodes. Figure 7.20 Context menu for links Linking multiple nodes simultaneously • After you cut all links, select three of your nodes by holding down the Ctrl key or by drawing a frame around the three nodes with the mouse cursor. • Click on the Link button in the contextual network ribbon. • Move the cursor to the target node (see Figure 7.21). The target node shows a green frame. Left-click and select a relation. The chosen relation is used for all links. If the relation does not apply to all links, you can change the relation in the next step via the link context menu as explained above.

210 QUALITATIVE DATA ANALYSIS WITH ATLAS.TI Figure 7.21 Linking multiple nodes The various ways of linking explained here apply to all entities that can be linked to each other in networks. When linking codes to codes and quotations to quotations, you are offered a list of link labels to choose from. All other links cannot be named. Adding nodes • You can add entities into a network editor via drag and drop from all managers, the Project Explorer or the margin area. Another option is to click on the Nodes tab and select Add Nodes. This adds a side panel to your network from which you can choose the various entities. • Click on the down arrow to select a node type. Then select the items to be imported and click on the Add button. You can also double-click to add an entity or drag and drop entities into the network. Figure 7.22 Adding nodes to a network

RECOGNIZING AND VISUALIZING RELATIONSHIPS – WORKING WITH NETWORKS 211 • The added nodes are placed in the upper-left-hand corner. To distribute them evenly in the network, select one of the Layout options. Removing nodes If you want to remove an item from a network, right-click on the node and select Remove from View or press Del. The delete key on your keyboard is the equivalent of the ‘remove from view’ option. Moving nodes If you want to move the whole or parts of the network within the editor, some nodes need to be selected first: • To select all nodes, use the key combination Ctrl+A. To select only a few nodes, hold down the Ctrl key. • Point to a node with the cursor and move the selected nodes by dragging the cursor to a different location. Accessing data behind nodes • Double-click a code node. You see the code comment, groundedness and density information and an option to show all linked quotations: List Quotations (Figure 7.23). This opens the list of linked quotes. This is the same behavior that you know from the Code Manager. Select each quote with one click to display it in the context of the data. If you dock the list to the right or left side of your screen, it will not be over other windows and will not get in the way. In version 8.4 or higher you can switch between list and preview mode. This means if the quotations are not too long, you can read them without having to load the documents to view the full quotation in context. Figure 7.23 Accessing the data behind code nodes in a network

212 QUALITATIVE DATA ANALYSIS WITH ATLAS.TI Layout options You can select from 12 automatic layout options. The results of the automatic layout procedure are typically quite usable and provide a good starting point for subsequent manual refinement of nodes’ placement. If you followed Skills training 7.2, you have probably already played with a few of them. The layout options can be combined with four routing options that are responsible for an optimal placement of the links. The following layouts and routing options are available: Figure 7.24 Layout and routing options for networks • You can access them in the Network Editor via the main Network tab and the View tab. You can use either the network from the sample project that you imported while working through Chapter 2 or one of your own networks that you created as part of the skills trainings in this chapter. The following is a description of how the layouts and routing options are calculated: The perpendicular layout allows the edges of the graph to run horizontally or vertically, parallel to the coordinate axes of the layout. All lines are at a right angle. It produces compact drawings with no overlaps, few crossings and few bends. The orthogonal tree layouts are similar to perpendicular layouts but process larger sub trees using a specialized tree-layout algorithm, which is better suited for tree-like structures. The circular layout places the nodes on circles, choosing carefully the ordering of the nodes around the circle to reduce crossings and place adjacent nodes close to each other.

RECOGNIZING AND VISUALIZING RELATIONSHIPS – WORKING WITH NETWORKS 213 The circular single-cycle layout is similar to the circular layout, but sub groups are not created and all nodes are placed on a single circle. This is useful for creating an overview and for shallow hierarchies. The organic layout is based on the force-directed layout paradigm. When calculating a layout, the nodes are considered to be physical objects with mutually repulsive forces, like, for example, protons and electrons. The connections between nodes also follow the physical analogy and are considered to be springs attached to the pair of nodes. These springs produce repulsive or attractive forces between their end points if they are too short or too long. The layout algorithm simulates these physical forces and rearranges the positions of the nodes in such a way that the sum of the forces emitted by the nodes and the edges reaches a (local) minimum. Resulting layouts often expose the inherent symmetric and clustered structure of a graph; they show a well-balanced distribution of nodes and have few edge crossings. The radial layout arranges the nodes on concentric circles. The layout calculation starts by conceptually reducing the graph to a tree structure whose root node is taken as the center of all circles. Each child node in this tree structure is then placed on the next outer circle within the sector of the circle that was reserved by its parent node. All edges that were initially ignored are re-established and the radii of the circles are calculated taking the sector sizes needed by each whole sub tree into account. This layout style is well suited for the visualization of directed graphs and tree-like structures. The hierarchical layouts aim to highlight the main direction or flow within a directed graph. The nodes of a graph are placed in hierarchically arranged layers such that the (majority of) edges of the graph show the same overall orientation – for example, top to bottom. Additionally, the ordering of the nodes within each layer is chosen in such a way that the number of edge crossings is small. Top–bottom. Prefers to place nodes downwards from top to bottom along directed links. Bottom–top. Prefers to place nodes upwards from bottom to top along directed links. Left–right. Prefers to place nodes from left to right along directed links. Right–left. Prefers to place nodes from right to left along directed links. The tree layout is designed to arrange directed and non-directed trees that have a unique root node. All children are placed below their parent in relation to the main layout direction. A child–parent relation in ATLAS.ti is defined via a transitive or asymmetric link. Before applying the layout, all symmetric relations are removed and added after the tree is laid out by connecting them with curved edges. Tree-layout algorithms are commonly used for visualizing relational data. The layout algorithm starts from the root and recursively assigns coordinates to all tree nodes. In this manner, leaf nodes will be placed first, while each parent node is placed centered above its children. Random layout. Randomly places the nodes every time this layout is invoked.

214 QUALITATIVE DATA ANALYSIS WITH ATLAS.TI Routing Orthogonal routing. This is an algorithm for routing a diagram’s edges using vertical and horizontal line segments only. The positions of the diagram’s nodes will remain fixed. Usually, the routed edges will not cut through any nodes or overlap any other edges. Polyline edge routing. Polyline edge routing calculates polyline edge paths for a diagram’s edges. The positions of the nodes in the diagram are not altered by this algorithm. Edges can be routed orthogonally, i.e., each edge path consists of horizontal and vertical segments, or octilinearly. Octilinear means that the slope of each segment of an edge path is a multiple of 45 degrees. Organic routing. This algorithm routes edges organically to ensure that they do not overlap nodes and that they keep a specifiable minimal distance to the nodes. It is especially well suited for nonorthogonal, organic or cyclic layout styles. Straight routing. Draw the links between nodes as straight lines without any consideration of node and edge crossing. The routing layout is not saved in the project. When you close a network, routing is lost. If you open a network in which you have applied routing, you must reapply it. Network view options Under the View Tab in the Network Editor, you can change the link name and switch between ‘Name’, ‘Short’ and ‘Symbol’. You can use short or symbolic names to save space on the screen. Another possibility is to use the short-name field not for a short name but for another language. For example, if you have data in English and Spanish in various research projects, you can use a Spanish word for the short name and use the short name in networks for all your Spanish-language projects. Figure 7.25 Network view options Other options are to hide or view comments, to display code-document connections, to preview quotation content and to hide or display frequencies or node icons.

RECOGNIZING AND VISUALIZING RELATIONSHIPS – WORKING WITH NETWORKS 215 SKILLS TRAINING 7.4 WORKING WITH THE RELATIONS EDITOR The Relations Editor gives you an overview of the existing relations and their properties. Additionally, you can define new relations. In the following, you will learn how to customize relations so that you can illustrate the kinds of relations that are relevant for your data material. Above, I introduced you to the concept of first- and second-class relations. You have seen that codes can be linked to other codes and quotations to other quotations via firstclass relations. For this process, ATLAS.ti offers a distinct set of link labels for each of the two entity types, as the nature of the relations between codes is different from that of the relations needed for quotations. The link labels offered for quotations include ‘discusses’, ‘justifies’ or ‘explains’ as compared to ‘is a’, ‘is associated with’ or ‘is part of’ for codes. Accordingly, there are two sections in the Relations Editor: one for code–code relations and one for quotation–quotation relations. The latter are called hyperlinks. In terms of handling, there is no difference in creating new code–code relations or hyperlinks. You just have to pay attention to which window you are in when you create new relations. Figure 7.26 The Relations Editor Opening the Relations Editor • If you are in a network, select the Relation Manager button from the Network ribbon. Or, from the Home tab, click on the drop-down button Links and select Relations.