0

I am hoping for some help with rank-ordered logit models. I get logit models, but Im having trouble when there are multiple levels, and when these are ordered. Moreover, I'm uncertain how to interpret the coefficients.

As an example, imagine you have a set of attributes for a car, and there are 6 characteristics that are ranked by individuals. To keep it simple, assume nothing missing, everyone has ranked each characteristic. So imagine I estimate my model, and have a series of covariates for the individual: age, sex, income, education. How do I interpret these coefficients with respect the ordered rankings? Also, can they then be converted into odds ratios (I dont see why they couldnt be, but like I said, this is new to me).

Moreover, if I just want to know what the relative importance of each characteristic is amongst my sample, how can I tell that? Is there a better way to do it?

Note: Bonus points if you can provide SAS or R code!!

Thanks in advance!

Rainmaker
  • 51
  • 6
  • Have you already looked up Multinomial and Ordinal regression? – Digio Feb 06 '18 at 15:48
  • Yes, Ive been able to run regressions in Stata. My problem lays more in trying to understand the estimated coefficients (using 'rologit' in Stata). I'm also not sure that I've set up the data correctly.... but, the 'terrifying' thing is that there *are* results. – Rainmaker Feb 06 '18 at 15:51
  • For starters read here: https://stats.stackexchange.com/questions/143024/how-can-i-explain-proportional-odds-models-to-a-layman/143026#143026 https://stats.stackexchange.com/questions/18119/how-do-you-predict-a-response-category-given-an-ordinal-logistic-regression-mode https://stats.stackexchange.com/questions/94581/plot-and-interpret-ordinal-logistic-regression and https://stats.stackexchange.com/questions/97582/linear-mixed-effect-model-vs-ordered-probit-vs-ordered-logit-with-ordinal-resp – AdamO Feb 06 '18 at 15:52
  • I found this article http://onlinelibrary.wiley.com/doi/10.1002/1521-4036(200010)42:6%3C677::AID-BIMJ677%3E3.0.CO;2-O/abstract helpful in learning about ordered categorical models. The code is SAS and SPlus but I think the other insights may help you. – mdewey Feb 06 '18 at 16:12

0 Answers0