It is known that conducting post-estimation tests for OLS assumptions (Multicollinearity, heteroscedasticity, and endogeneity) is necessary.

But is it statistically necessary to carry out these OLS assumption tests for other models other than multiple linear repression model, like for Tobit model?

  • $\begingroup$ 1. Usually formal testing of assumptions is less useful than simple diagnostic checks (it answers the wrong question). 2. OLS and Tobit share a few assumptions. But in respect of other aspects of the models, they make quite different assumptions -- it would not be appropriate when doing Tobit to check an assumption that Tobit doesn't make. $\endgroup$ – Glen_b -Reinstate Monica Feb 3 '15 at 21:37

Normally it makes sense to check whether the assumptions that you are building on are satisfied. If they are not, then you cannot trust the results you got since their validity depends on the assumptions being satisfied. Hence, if you do OLS estimation, you would check OLS assumptions. If you do other kinds of estimations, you would check the relevant assumptions (which need not be the same as OLS assumptions).

| cite | improve this answer | |
  • $\begingroup$ Thank you Richard Hardy, what I am confused of is that some of the researchers suggested that OLS assumptions must also be satisfied for other estimations (like in the case of Tobit, probit/logit, Hechman two-stage model). Anyways from your response I understand that it makes non-sense to check OLS assumptions for other than OLS estimations. Currently I employed Tobit model for analysing some problem, so may you suggest me some relevant assumptions to be checked for Tobit model. Thank you $\endgroup$ – Bizualem A Feb 3 '15 at 10:57
  • 1
    $\begingroup$ There may be cases where the same (or a similar) set of assumptions underlie more than one model. It would not change anything if you used different names for the same assumptions (although this could easily introduce confusion). Sorry, I have never really worked with Tobit; I hope someone else can help you with that one. $\endgroup$ – Richard Hardy Feb 3 '15 at 11:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.