Hi, We have collected an open-ended answers to question to estimate people's affective state. We then ran a content analysis, following some pre-defined affective categories. In a result we have 3 categories, which we would like to use in quantitative analysis - particularly, looking at differences between people in each category on our continuos dependent var (probably running ANOVA or Kruskal-Wallis). My question is, is it correct to use such categories as independent variable?
I don't see any problem with coding an open-ended variable, and then using the resulting data in quantitative analysis. Many qualitative software packages (e.g., MAXQDA) refer to this as a form of mixed methods research, but I think that goes too far. At most, it amounts to quantifying qualitative data.
I don't see any problem with coding an open-ended variable, and then using the resulting data in quantitative analysis. Many qualitative software packages (e.g., MAXQDA) refer to this as a form of mixed methods research, but I think that goes too far. At most, it amounts to quantifying qualitative data.
Yes, you can use these as independent variables, but you obviously need enough respondents/participants to run these tests and find some significant differences!
David L Morgan and Rodney Duffett thank you for your answer. Also, we have relatively large sample size :)
Thanks again for confirming. Would you happen to know any literature that discusses such data analysis? I have found many sources about mixed method but they don't really mention the use of categorical 'qualitative' results as independent variables.
The use of open-ended questions as sources of code-able variables has a long history in survey research, where it is considered to be a standard practice.
Hi, I did the same in one of my studies. I followed an inductive approach with qualitative data to study the work life balance of female lecturers in private higher education institutes in Kandy. And with thematic analysis I identified the factors affecting work life balance of them. I recommended under 'Further Study', to conduct a quantitative analysis to prove the relationship in between independent variables (the themes identified) & the dependent variable.
Thank you for your answer and suggestion Niluka Samarakoon I really appreciate it.
Yes, your case is a qualitative -quantitative approach, the only caution is that you have to ensure the samples were random enough to allow you to run ANOVA. ANOVA is not recommended since the data may have been derived from purposively selected interviewees. I recommend you use Kruskal-Wallis instead of ANOVA. The Kruskall-Wallis does not have the underlying assumption of data normality.
Joseph Kipkorir Cheruiyot Non-random samples are used all the time in experimental designs (the data are often randomly assigned but from purposively selected groups). Also, random sampling is not the same as data normality.
More generally, in converting open-ended questions into content-coded variables, there are usually two options. One is to create dummy variables, according to whether or not something was mentioned. The other is to count the frequency of mentions.
@David L Morgan, thank you for your clarification Prof. Supplementary question; did you refer to the above case as experimental design? Seems to have qualities of an ex post facto design.
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running