Diagnostic tests in a ordered probit

[Cambridge101]

Hi there,

I am estimating an ordered probit model and was wondering which diagnostic tests are possible using this model. I haven't been able to find any information or commands for:

- heteroscedasticity
- goodness of fit
- normality
- functional form

or any other diagnostic tests that anyone can recommend.

Any help would be much appreciated!

Thank you

[Richard Williams]

You can test the parallel lines assumption (which, if you were doing ologit, would also be called the proportional odds assumption). Download -gologit2- from SSC. Then do something like

webuse nhanes2f, clear
gologit2 health female black age, pl sto(oprobit) link(p)
gologit2 health female black age, npl sto(goprobit) link(p)
lrtest oprobit goprobit, stats

For more, see

http://www.stata-journal.com/article...article=st0097

http://www3.nd.edu/~rwilliam/gologit2/index.html

[Richard Williams]

As for hetero, you can check out

http://www.stata-journal.com/sjpdf.h...iclenum=st0208

I'll note, however, that there is a lot disagreement over how to deal with hetero. For a short summary see

http://www.europeansurveyresearch.or...W_ESRA2013.pdf

[Cambridge101]

Dear Richard,

Thank you for your suggestions. I have tested the parallel lines assumption using lrtest and the brant test.

I'm still looking for other diagnostics. Unfortunately the above do not show how to explicitly test for hetero.

Thanks

[Bui Thanh Yen Thao]

Hi statalists,

I also face the problem: the parallel regression assumption for the ordered probit model is violated. It is usually advised that we should alternate other possible models: multinomial logit model, generalized ordered logit model.

However, the research that I refer to for my thesis also analyze this problem and still use ordered probit model although its parallel regression assumption is violated. The drawback of using the multinomial logit model is that it does not preserve the inherent ordering of the categories of dependent variable (6 values) and therefore does not incorporate this information when estimating the coefficients of the explanatory variables. This results in a loss in the efficiency of the estimators (Long, 1997). While the generalized ordered logit model provides an alternative model that does preserve the ordering (e.g., it is a restricted version of the multinomial logit model), it is very sensitive to low frequency counts (e.g., small cell sizes). Thus, it is often necessary to combine the dependent variable categories that have low frequencies with adjacent categories in order for the estimation procedure to work. However, combining categories may also lead to a loss in information, especially if the underlying latent variable is multi-leveled or continuous. As a result, we have chosen to present the results from the ordered probit model. A larger sample size and fewer explanatory variables would have made the use of generalized models more feasible.

I wondered what happens if we do like that. Could you please help me to understand this situation? Thanks.

[Bui Thanh Yen Thao]

I am a new-comer in this field, specially with ordered probit model. However, this is the problem for my thesis. Pls help me!