How to use ranked data from survey

[Imrul Huda]

Hi!

I asked my participants to rank the 03 most important factor they look for in a job. Basically I gave them 09 options and participants picked 03 most important and also ranked them 1,2,3. I think the right term for this is incomplete rank data. I have done my EDA with tables and graphs. Now I want to use this data to explain some other variables. Such as "earnings".

I also had a likert scale like question that I used to conduct a factor analysis. I can use the factor scores for classification, regression etc. I am looking for something similar for my ranked data.

Thanks!

[Hong Il Yoo]

You can use -rologit-, which is a rank-ordered data version of -clogit-. If your final plan is to estimate models that capture unobserved preference heterogeneity, you may find it more useful to "explode" your rank-ordered response into "pseudo-choice" responses, and use -clogit- on the exploded data to replicate what you can get from using -rologit- on the original data. You can check out section 6 of this article (https://journals.sagepub.com/doi/pdf...36867X20931003). You can then use -mixlogit-, -gmnl-, -lclogit2-, and the like on the exploded data to estimate a rank-ordered logit model with unobserved preference heterogeneity.

[Imrul Huda]

You can use -rologit-, which is a rank-ordered data version of -clogit-. If your final plan is to estimate models that capture unobserved preference heterogeneity, you may find it more useful to "explode" your rank-ordered response into "pseudo-choice" responses, and use -clogit- on the exploded data to replicate what you can get from using -rologit- on the original data. You can check out section 6 of this article (https://journals.sagepub.com/doi/pdf...36867X20931003). You can then use -mixlogit-, -gmnl-, -lclogit2-, and the like on the exploded data to estimate a rank-ordered logit model with unobserved preference heterogeneity.

Thanks! Can you please briefly explain what you mean by "explode"?

[Hong Il Yoo]

Please see section 6 of the hyperlinked paper.