[R] Distribution, heteroskedasticity etc. tests for glm, tobit and heckit?

Michal Kvasnička prgosek at gmail.com
Wed May 29 08:29:55 CEST 2013


Is there any package / code snippet to test the distribution
assumption, heteroskedasticity, omitted variables, and linearity with
the models estimated by maximum likelyhood? I especially need it for
three type of models:

* binary choice (probit and probit with non-normal distribution) --
estimated by glm
* tobit and tobit with non-normal distribution -- estimated by AER or
any other suitable package
* heckit and heckit with non-normal distribution

To be more precise, when I estimate e.g. probit model, I should test
that the random part in the data/model is normally distributed. If it
was not, the parameter estimates may be biased. In the help and books
I found how to estimate the parameters but not how to test the
distribution assumption (and other assumptions stated above).

Is there any package or code snippet, or do I need to derive the test
for every distribution and model, and code it myself?

Best wishes,

More information about the R-help mailing list