An exploratory survey asked human trafficking aftercare workers:
What are the religious beliefs of the trafficked persons you/your agency service? [Mark all that apply]
- Buddhism
- Confucianism
- Taoism
- Ancestral worship...
How do trafficked persons understand the reasons for the violence they have suffered? [Likert scale response: Never/NA, Seldom, Sometimes, Often, Almost Always]
- They have done something wrong
- God's punishment
- God is unfair
- Karma
- Fate
- Bad luck...
I am searching for the best statistical model (for a non-statistician) to use to answer questions such as: Are Buddhists more likely to hold one belief than another? Are Taoists more likely to believe that God is unfair? than Buddhists?
N = 53
I am using SPSS (mac) V22.
Generally, Likert-type survey results are assumed to be ordinal (there is some debate on this, but generally comes from the misunderstanding of the difference between Likert-type and Likert scale surveys; see: http://www.joe.org/joe/2012april/tt2.php ). With that said, you'll need to do non-parametric analysis. Chi square using counts of your results can answer your questions easily and is very approachable for a non-statistician. Alternatively, you may explore correlation using non-parametric techniques (Kendall tau). Good luck with it!
Hello Suzanne,
Chi-square test is one of the solutions. However, there is one serious drawback. If you have numerous categories (3 or more) chi-square is indicating the significant differences but does not say where exactly they are.... say is it between 1 and 2 or 2 and 3 or 1 and 3 categories.
But on the other hand I am not sure if something much better than chi-square would be applicable here.
Best wishes
Dear Suzanne,
I'm afraid I disagree with the other two responses and would not recommend chi-squared because you are doing multiple answer multiple choice leading to repeated measures. You need to turn your data into binary (0 or 1) variables for each possible answer. I presume you wish to order the responses for each question. The best test for this is Cochran's Q which works on repeated binary data.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling