Analyzing ranked data in Stata 15

[Seth Meyer]

Hello everyone,
I have survey data of 1000 people who ranked choices with what qualities are most important for a mental health counselor. They had choices like sexual orientation/gender identity, race, location, specialty, etc. Because I collected the data on qualtrics, the data is set up with each choice (e.g. location) has how many people chose it as 1st, 2nd, 3rd, etc. What I would like to do is analyze the differences between subgroups in what is important. I have Stata 15. Are there ways to create models to show how different groups rank choices differently or to analyze ranked choices in Stata 15?

Thank you for your help.

Seth

Q22_1 | Freq. Percent Cum.
------------+-----------------------------------
1 | 91 12.04 12.04
2 | 77 10.19 22.22
3 | 118 15.61 37.83
4 | 229 30.29 68.12
5 | 185 24.47 92.59
6 | 56 7.41 100.00
------------+-----------------------------------
Total | 756 100.00

. tab Q22_2

Q22_2 | Freq. Percent Cum.
------------+-----------------------------------
1 | 355 46.96 46.96
2 | 200 26.46 73.41
3 | 108 14.29 87.70
4 | 56 7.41 95.11
5 | 28 3.70 98.81
6 | 9 1.19 100.00
------------+-----------------------------------
Total | 756 100.00

[Richard Williams]

Stata 15 has rologit, later replaced by cmrologit. For a discussion of rologit models, see

Quiroz, C., & Williams, R. A., (2020). Rank-Ordered Logistical Models, In P. Atkinson, S. Delamont, A. Cernat, J.W. Sakshaug, & R.A. Williams (Eds.), SAGE Research Methods Foundations. https://doi.org/10.4135/9781526421036940688

Excerpt:

The rank-ordered logistical model (commonly abbreviated as ROLM or ROLOGIT) is used to analyze data in which respondents are provided a preset number of options and are subsequently asked to place those options into a complete order of preference. ROLM is a generalization of the conditional logistical model (CLM) for ranked outcomes (Long & Freese, 2014). CLM differs from ROLM in that CLM is utilized when individuals provide data only on their single best outcome. For example, if an individual were given three options—bus, car, or train—and asked which single mode of transportation they preferred, then these data collection technique would call for a CLM (see Long & Freese, 2014). However, if the same individual were asked to rank each of the options in a complete order of preference—first, second, and third—then the greater amount of information provided would call for an analysis using ROLM.

The ROLM can also be used to determine if characteristics of the ranking individuals play a significant role in how ranks are ultimately decided. Suppose one were interested in the various determinants of how job applicants were ranked. An ROLM would let one examine the impact of education and job experience on an applicant’s rank. The same model would also be used in order to show how the raters’ characteristics affected the rank of the applicants. For example, one could demonstrate whether an applicant’s job experience was a more important factor for female raters than it was for male raters.