3
$\begingroup$

I've got a question concerning the estimation of a tobit model with the AER package in R. I observed t-distributed residuals, so that the assumption of normal distributed std errors is violated. Fortunately it's possible to choose "t" as the distribution when fitting with tobit(). It is tobit(<formula>, dist="t"). After changing the distribution from t to normal, the AIC increases, so this seems to be a good idea.

My problem is: Is it possible to flip the normal distribution that way? Is it still a tobit model and do I have to care about something when changing the distribution in a tobit model?

I appreciate any thought!

$\endgroup$
0

1 Answer 1

0
$\begingroup$

Have a look at this article, which may be helpful for answering the question: "Student-t cenbsored regression model: properties and interference"

Arellano-Valle, R.B., Castro, L.M., González-Farías, G. et al. Student-t censored regression model: properties and inference. Stat Methods Appl 21, 453–473 (2012). https://doi.org/10.1007/s10260-012-0199-y [Titel anhand dieser DOI in Citavi-Projekt übernehmen] https://doi.org/10.1007/s10260-012-0199-y [Titel anhand dieser DOI in Citavi-Projekt übernehmen]

$\endgroup$
1
  • $\begingroup$ Hi. Welcome to CV. While this might help, it is appreciable if you could provide the gist or the relevant materials from the papers you linked for link only answers are prone to get useless once the links die. $\endgroup$ Oct 29, 2022 at 17:36

Your Answer

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