Ranked Data analysis Help

#1
Hi
I need help analyzing data from a survey. i am looking at consumers shopping venue and their confidence in regulation by various industry players.
Shopping venues question: please choose your top 3 shopping venues (1 for each and zero for those not chosen)
Confidence questions: How likely are you to trust information for the following sources: ranked from 1 to 5 (most likely to least likely)
I used a Logit model to analyze the data and the results look fine. However i am wondering if given the ordered nature of the data i can be able to use a different model that will better suited the situation.
I am using STATA for my analysis
Also keep in mind that causality is not determined so the regression can be runned both ways
 

terzi

TS Contributor
#2
Hi nickys,

Logit, ordinal or rank ordered logit models can be suitable, but I'm not quiet sure if it's correct to apply each of those models changing the place of response and predictors. I'd recommend some descriptive analysis to be performed before the model (it is also easier, so if you don't find any indications of relationships you probably can just quit there). I would suggest starting with some descriptive analysis that would yield to some possible relationships. For instance, you could perform a correspondence analysis to check whether people shopping in one particular store tend to relate with some levels of trust information.

Once you have some insights about the relationships you may be looking for you can suggest a better model and even code the predictors accordingly (the coding used in predictor variables can produce different results).

Good luck
 
#3
Hi Terzi,
Thank you for your response,
Just to make things clear, i did run a Logit model and there a clear relationship between the certain types of venues and confidence. However i was thinking that given the nature of the data( shopping venue as top 3(1 for the top 3 and 0otherwise)) there could be another type of model that will suite the data better. Since the dependent variable is not ranked it is not possible for me to use the ordered Logit. I think my biggest problem here will be proving causality( do respondent shop at location Y because they have confidence in X or is it because they have confidence in X that they shop at location Y...
Thank you