Percentage Prevalence comparison (non-independent groups)

#1
Hi all!

I have a dataset which I think could potentially be very exciting but I am struggling to find how to analyse it!

We have gathered lots of information on the percentage if individuals who exhibit different forms of the same class of behaviour (categorical data- present/absent). All individuals have the potential to engage in multiple forms, therefore the groups are not totally independent. Furthermore, unfortunately we are not able to say for a given individual exactly which behaviours they exhibit- the datasets for each form of behaviour aren't linked (however they were mostly the same individuals assessed for the different forms of behaviour). I want to compare the prevalence of the forms of behaviour to see if one form is more less common than another.

Due to the groups not being independent, I am unsure whether a chi-squared test is appropriate. However, preliminarily I entered the data into SPSS coding 1=present, 2=absent in the first column then in the second column coding the forms of behaviour as 1, 2 etc and it came out as being significant. However, obviously that is entering them as if they are independent and I am concerned that it is too much of a fundamental violation of the assumptions of the test.

As an example- we have 745 participants: 198 engage in form 1, 401 in form 2, 243 in form 3 but for participant X we cannot say yes1, no2, no3 (sadly :()

I am unsure whether there is a test which I can do to compare percentages? Is it possible to use a t-test in this way?
 
Last edited: