Author: Áine M. Humble

Analysis processes (continued)

Meadows and Morse (2001) note that different combinations of "analytic approaches (e.g., constant comparison, immersion/crystallization, matrices, manual analysis, and computer-assisted analysis)" (p. 194) can enhance the rigor of a study. I would agree with this - I came to what I felt was a robust and valid analysis only through using several data analysis techniques (note: these are described in more detail in Humble, 2009 [PDF- 139 KB]; Humble, under review). I have already described some of the techniques in the third section where I described the initial analysis; here I describe how I extended that analysis.

I should point out that each technique, both in the initial and later stages of data analysis was important in drawing my attention to different features of wedding planning. Like the different lenses an optometrist places in front of a patient, each brought the full phenomenon of wedding planning into clearer focus. Additionally, using different data analysis techniques helped me to become aware of blind spots I was holding with regard to the analysis. Ultimately, through the process of carrying out these different ways of looking at the data I developed a more complete and validated picture of the process of gender construction within wedding planning for these couples.

Rank order comparison
Rank order comparison (Curtis et al., 2001) involves the counting of certain codes. Counting can occur in qualitative research and in fact some researchers argue that it is an intrinsic part of qualitative work (Miles & Huberman, 2004; Sandelowski, 2001) despite the ways in which many researchers attempt to "mask" numbers with words (Sandelowski). If counting is used, it should contribute something further to the qualitative analysis, such as enhancing the trustworthiness of an analysis (Miles & Huberman) rather than, of course, replace it.

In this study, I counted codes for the number of different types of wedding work individuals reported (a) women doing on their own, (b) men doing on their own, and (c) couples carrying out together. Based on these counts, I came up with an average number of tasks performed by women, men, or together within each of the couple categories and then the couple categories were ranked compared to each other. This procedure is a bit too detailed to describe here, but is described in Humble (under review). The important thing to note here is that it was only during this analysis that I had an "a-ha" moment in which I realized that I had been focusing too much on women's solo work or the work couples did together at the expense of the work men did on their own (my first blind spot). Examining the rank ordering of the different couple categories in terms of who carried out what work, I realized that the one way that transitional couples differentiated themselves from egalitarian couples was that men in transitional couples rarely did anything on their own, in fact they reported doing even less tasks on their own compared to traditional couples in this study. I would not have had this awareness without conducting the rank order comparison. Enhanced clarity!

I mentioned earlier that two features of the MAXQDA program assisted with carrying out the rank ordering comparison. The first feature was the Attributes feature, a spreadsheet option in which descriptive coding (Richards & Morse, 2007) that stores information about each file (in this case, each person) can be entered. Once descriptive codes or "attributes" are entered, transcripts can then be "activated" on the basis of the values of that attribute. For example, if I was only interested in how gender affected wedding preparation, I could activate the files by two values (women or men) located in the gender attribute. Other examples of attributes I entered were whether or not the couple had cohabited prior to getting married and who was remarrying (the bride, groom, or both). Some attributes I entered prior to data analysis, but I also added attributes that emerged out of the analysis. For the purposes of the rank order comparison, I added in which couple category each person was assigned to (traditional, transitional, or egalitarian). Once I had this in the Attributes table, I was then able to activate files according to their gender construction (note: such activations can actually be saved as sets, which can save time further down the road) so that I could then look at coding for traditional, transitional, or egalitarian couples.

Setting up the activation of the gender construction code paved the way for me to use the Code Matrix Browser option, which provides a visual look at which codes (all codes may be looked at or only certain codes) have been assigned to which transcripts (a second option is to look at how many times a code has been assigned to a particular transcript, but in this case, I was only interested in the presence or absence of the codes). As part of the rank order comparison, I activated the categories of the three types of work (work done by women, by men, and together) and analyzed them one at a time by each couple category. An example of this is shown on my webpage.

Visual analyses
The second angle I took on the wedding work involved an examination of respondents' TextPortraits. TextPortraits are a visual representation of coding, available in the MAXQDA program. To use TextPortraits, different colors must be used for different codes. In this case, I coded women's work in yellow, men's work in blue, and work done together in red. Additionally, white was used to code for specific comments referring to lack of help from others. I then created a TextPortrait that looked specifically at these codes. A TextPortrait is a within-case display (Kuckartz, 2007a), meaning that it looks at specified coding for just one individual (or case) at a time. TextPortraits are described in the following way:

Starting with the first row and the top left tile, the tiles are "described" with a color from left to right. When it reaches the end of a row, it proceeds down to the first column of the following row - much like when using the "return" function on a typewriter. In the normal manner of presentation, the total number of tiles is divided among the coded passages of text so that the number of tiles that symbolize a segment or its color is determined by the total size of all coded segments. (Kuckartz, 2007a, p. 8)

Thus, "broken down into a 100 X 100-piece rectangle, the (TextPortrait) is 'portrayed' as a painting of its codes, maintaining the location and length of their occurrence in the text" (Kuckartz, 2007b, p. 12).

At this stage I took another break from working at the computer. Similar to what I did in the initial analysis, I took my analysis to the floor again. I added each bride and groom's TextPortraits into the word document that I had already created, the one containing a paragraph description of each couple's wedding, and these sheets were printed out in color so I could effectively examine the TextPortraits. I placed these on the floor and looked at them, considering the range of gender construction and what I was seeing in the TextPortraits.

After I examined these visual representations of selected codes, I realized how important it was to activate the "correct codes" or else the TextPortraits could be interpreted incorrectly. When I originally created the TextPortraits for each participant, I only included the three categories of work previously described (women's work, men's work, and work done together). However, as I looked at the portraits on the floor, one man's portrait, in particular, made no sense to me and made me question my initial categorization of him and his wife. I realized I had made an error in my set up of the visual analysis of couples' wedding planning. I had neglected to consider how other people's contributions (e.g., female friends, mothers, fathers) - my second blind spot-influenced the overall proportion of how much brides and grooms contribute to wedding work. When other people's contributions were not included in the TextPortraits, some of the men's wedding work was misrepresented (specifically, given greater presence than it should have) in the pictures. As a result, I printed out two different TextPortraits for each individual - the first one consisting of four activated codes (women's work, men's work, work done together, and lack of help from others) and the second portrait including other people's contributions (represented in grey color). Examples of two of these TextPortraits are shown at my website and explained further there. I took the 14 sheets back to the floor a third time for examination, comparing TextPortraits within individuals, within couples, and between couples. Using the TextPortraits helped me develop a greater understanding of what was going on with the findings, and showed me that data should constantly be reviewed at all times in a project to ensure rigor (Morse, Barrett, Mayan, Olson, & Spiers, 2002).

Learning and using software
Despite knowing how to use the general features of MAXQDA software prior to beginning this study, I should point out that the Code Matrix Browser and TextPortraits were new options available in the program that I was not familiar with. I learned how to use these new tools as I carried out my study; these tools were explored after I had completed the initial data analysis described in the third section. Learning the visual tools as I was conducting my analysis worked well for me, but for others it might be imperative that they be familiar with all aspects of a software program before beginning their study or analysis. Indeed, during this study I realized that I had not used the Attributes function to its full capabilities (to create "sets") in past research. When I did finally realize how I could use this tool, I was very excited, and it greatly assisted with using descriptive coding (Richards & Morse, 2007) to explore additional patterns in the data. Being familiar with the main characteristics of a software program is vital before starting any analysis.

However, I do want to make a final point about using QDA software. I am a great fan of QDA software for assisting with data analysis (as we all should know, software programs do not perform the analysis or interpretation). I have never conducted qualitative research without it (so I can not compare how my analysis might be different if I was doing it by hand) and I probably will not (however, not all of my analysis takes place within the software program either). But I can certainly see that the development of new tools in QDA software changed the way that I looked at my data. The capabilities of QDA software open up so many doors for data analysis (Kuckartz, 2006, 2007, has reported on this topic), but suddenly the question is raised for me - did I come to mistrust my initial analysis as a result of the new functions in MAXQDA or did I realize that I could have done more with my original research? I do not have an easy answer to this interesting question. I look forward to seeing more discussions around QDA software, and I note that a list of papers addressing QDA software issues is already available in the bibliography section of The CAQDAS Networking Project.

I know that one reason I did not use any visual techniques for data analysis in my first study on wedding preparation was because the software did not allow it. At that time I used a format of the software that was two versions earlier than MAXQDA 2007 (for those who are interested, MAXQDA provides a history of its software). But I also did not do the rank order comparison (Curtis et al., 2001) the first time and yet I could have done it by hand in the first study. In fact, the first time I did it for this study, I did do it by hand, and it took me about five hours to complete it. When I realized I could have used the Code Matrix Browser for the same analysis, I did do it for verification, and it only took a half hour! And how did I learn about the rank order comparison technique? I first read about it in an article on content analysis (Hsieh & Shannon, 2005). I also was teaching a qualitative research course at the same time I was working on this project, and I learned about matrix analysis (Averill, 2002; Miles & Huberman, 2004) through this route. Whatever pathways a person takes, I can not emphasize the need to read, read, and read about qualitative research. The more I learn, the more I realize there is so much more to still know. And so I seek out more information through exploring software options, through basic "trial and error", talking to others, and by reading qualitative method books and journals (e.g., Qualitative Health Research). These strategies are not used in a particular order. It is a flexible and exciting process!

Back to Project Home Page