What to consider when using an ordered logit/probit regression?

Hi,
I'm studing the effect of working hours on self-reported health where health is scaled 0-10 and working hours is the difference between actual hours and preferred hours. Based on this I created two dummies for working more than prefered and one for working less than preferred. these two dummies is what I want to estimate on self-reported health. I will offcourse include age, gender, education etc as independent variables also. I will make groups of variables, age education etc.

To do this regression I will use polr in the MASS package.
But before or after running the regression what are those steps (if any) I sould code or consider according to econometric theories? For instance there is so much to do/think about when using OLS but I can't find any other about polr/ordered regression.
In other communities/forums it was suggested to use brant or anova test for ordered regression, so I will do that. But is this enough? I don't feel it's enough.
What about heteroskedasticity, endogeneity, robustness, causality? I'm very much in doubt about what to do.
Can you suggest any kind of paper, book etc. I can read according to the above?

Because these questions are not specific for the R environment, you might have more success by looking at the Cross Validated website.
But maybe that was one of the 'communities/forums' you already consulted.

Basically, the issues that you identify are the same for probit/logit as they are for OLS. The important one is likely endogeneity. And that's the one that's hard to solve, of course.

@HanOostdijk Thanks I will do that from now.. One of the reason that I post it here was that I don't now how R can be used for my purpose.. so I thought if I post here then maybe someone could give me answer with some r codes or packages.

@startz oh okay.. I was just very much in doubt because I could really not find any examples with it.. seems like no one considered one of these things when I search.

7 Proportional Odds Logistic Regression for Ordered Category Outcomes | Handbook of Regression Modeling in People Analytics: With Examples in R, Python and Julia (peopleanalytics-regression-book.org)

2 Likes

Do anyone know if the polr function in the MASS package is better than oglmx package? In the oglmx package there is heteroskedastic ordered probit regression which sounds better but I want to hear oppion and why to use which?

This topic was automatically closed 42 days after the last reply. New replies are no longer allowed.

If you have a query related to it or one of the replies, start a new topic and refer back with a link.