[R] PEA and APE Tobit

[Arne Henningsen]

Dear Annelies

On 26 February 2015 at 09:12, hnlki <annelies.hoebeeck at ugent.be> wrote:
> I estimated a tobit model
> tobit.fit<-tobit(y~x,left=0, right=Inf)  (library "AER")
> or
> tobit2.fit<-censReg(y~x, left=0, right=Inf) (library"censReg")
> I' have estimated the partial effect at the average as:
> pea<-(pnorm((colMeans(x)%*%tobit.fit$coef[-1])/tobit.fit$scale))%*%tobit.fitt$coef[-1]
> and the average partial effect as:
> ape<-
> (length(x[,1]))^(-1)*sum(pnorm((x%*%tobit.fit$coef[-1])/tobit.fit$scale))*tobit.fit$coef[-1]
>
> I guess I did something wrong as
>  margEff( tobit2.fit) (library("censReg")
>  gives a different result than my partial effect at the average.
>
> Any ideas about what I did wrong?
> I  did not find the underlying code of margEff.

[...]

> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

Please follow the posting guide and provide a self-contained
reproducible example.

Best regards,
Arne

-- 
Arne Henningsen
http://www.arne-henningsen.name