 |
 |
At 13:53 05/03/2004, you wrote: >However, I would have to disagree with Thomas when he wrote: > >>QDA Miner does not explore relationships; neither does any other CAQDAS. >>You will need to come up with some hypotheses about these relationships, What I really meant by this, was that you need to have an approach to how you want to explore an relationship. You cannot dump the data into any program, then press a button and expect a good analysis. No software, SPSS included, can do that. There are many good things about letting the computer to churn out word clusters the way you and Bob have mentioned. In fact, I was surprised, how useful Leximancer was in automatically generating clusters from completely unstructured data and I guess QDA Miner will be equally, if not more, useful for that. However, when I was checking Leximancer, I had already idea, why I wanted to get word clusters. If for my type of analysis word clusters or code clusters would have been meaningless, then there would be no point in performing such an analysis. I may have been presumptous here (apologies to Ian for that), but I had the feeling that the original poster was not very decided about his methodology and I would not want software, no matter how good or useful, to guide one's methodology. Having said that, I have to say that I am very content with the new mapping techniques. Here is why: What I did was that I identified five "frames" (Goffman/Gitlin) in my data, first by reading through part of the data. I then identified a number of keywords for these frames and coded these keywords and their lemmata. I then took the code matrix and churned it through all sorts of clustering techniques (k-means, factor analysis, latent class analysis). Sure enough, I got decent model fits (for LCA) and the hypothesized clusters. But then I churned the wohle data through Leximancer, and saw a sixth cluster, which I completely overlooked in my analysis, but which I could easily interpret. Thomas