Comparing frequencies
Hi JASP Team!
I have a dataset of 95 children who provided answers to an open question. I coded the answers and ended up with 11 categories which are not mutually exclusive (meaning that each answer can fall on more than one category). Now, what I want to do is to compare the frequencies of all categories with one another to find statistically significant differences. For instance, 43 children refer to height, 39 to body and 47 to style. Are these differences between the frequencies statistically significant? I am open to using Bayesian statistics, if you think that this makes sense.
I just downloaded the newest version of JASP 0.19.1.0, if this is important.
Many thanks!
Georgios.
Comments
Hi. I don't think you'll find a standard, straightforward way to do inferential statistics on the frequencies of NON-mutually-exclusive categories.
You could filter your data to include only those children whose answers fell in one and only one of the 11 categories. You could then to a set of chi square analyses to look for significant (or Bayesian-substantial) differences in frequencies.
I can imagine a generalized linear mixed effects model in which 'category' is a within-subject variable and response magnitude (either "1" when the child made a response in a particular category, or "0" otherwise), but I'm not sure how persuasive the result would be.
R
Thank you for your answer!
I thought more carefully about which are my research questions and organized the 11 categories into two super-categories. And then I created a three-value variable, with 1 (answers that fall only in the first super-category), 2 (answer that fall only in the second super-category) and 3 (answer that falls in both super-categories). So now the categories are mutually exclusive.
In this case, is the multinomial test the one I am looking for?
Yes, that's right.
R