1
$\begingroup$
  1. I would like to know if we need to test the assumption of residuals' normality when we have probit model? And if this assumption is violated how can I correct it with Stata?

  2. In the case of Probit model, how can I calculate the pearson matrix using Stata?

$\endgroup$
  • $\begingroup$ Be aware that CV is not a statistical software support site. Your question about the probit is on-topic, but Stata code may not come with the answer. $\endgroup$ – gung - Reinstate Monica Mar 23 '15 at 12:51
  • $\begingroup$ How do you define the residuals in a probit model? $\endgroup$ – Xi'an Mar 23 '15 at 17:10
0
$\begingroup$

1) Probit is part of family of generalized linear models (GLM). You can use deviance as a analysis of model fit.

http://en.wikipedia.org/wiki/Deviance_(statistics)

2) Do you mean covariance of regressors? If that is so then you need an information matrix..

$\endgroup$
  • $\begingroup$ Thank you for your answers. But i did not understand why i must use deviance. You said that probit is part of GLM, so it is necessary to test for normality. Regarding my second question I mean the "pairwise correlation", how can i calculate it for probit model using stata? Tahnk you $\endgroup$ – Rourou29 Mar 23 '15 at 15:35
  • $\begingroup$ @Rourou29 There is no residuals in least squares sense. Deviance is only statistic which act as one in GLM setting. Normal distribution is part of the picture only because your link function is probit. I do not understand what you mean by pairwise correlation. Do you mean between two regressors or what? $\endgroup$ – Analyst Mar 24 '15 at 10:18
  • $\begingroup$ Thank you very much. Yes I mean between two regressors. $\endgroup$ – Rourou29 Mar 24 '15 at 20:43
  • $\begingroup$ @Rourou29 you need an covariance matrix of regressors. Here is link to function stata.com/manuals13/pmatrixget.pdf $\endgroup$ – Analyst Mar 25 '15 at 6:17
  • $\begingroup$ @Rourou29 You are welcome. Perhaps you can select my answer as the best one? :) $\endgroup$ – Analyst Mar 26 '15 at 6:07

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.