Hi all-
I have a multi-item survey in which respondents marked "yes" or "no" to engaging in various activities. I'm hoping to compare the rates at which folks endorsed engaging in activity X (e.g., gardening) relative to engaging in the various activities W, Y, and Z (e.g., bowling, cooking, hiking). I'm looking for a statistical test of this difference.
I've computed the average rates of "yes" responses. But, I'm a little stumped on how to actually test whether the "yes" responses for X is more common. Since individuals can respond "yes" to multiple categories, I don't think a chi-square test is appropriate. I was thinking a multilevel model might be best, such that nesting within respondents could be accounted for - perhaps multilevel logistic regression?
Hi Simon, interesting question. If you are interessted in general activity patterns of your folks then maybe a latent class analysis of the overall response patterns and the differences between them may be also something I would consider.
If you want to see how the engagement in activity X can be prediced (depends on) the engagement in W, Y, Z etc., then logistic regression* is the tool of choice. It allows you even to look at interactions (e.g. a person may be engaged in gardening if he is engaged in cooking, but only if he is not engaged in traveling, too)
--
* a generalized linear model of the logistic family of distributions.
Dear Simon,
I suggest that you need to go beyond calculating yes/no frequencies as it the interaction between responses that counts. The most appropriate technique is Cochran's Q. This is available in SPSS under Analyze > Nonparametric tests > Repeated measures.
I think the most appropried test is Crocran's Q like Peter told you. Also you can do Rasch Analysis, is another analysis technique (I attached 3 references).
- Badges
- Science topic
- Similar topics
- Psychology