Message-ID: <alpine.DEB.2.20.1605292117280.791@paninaro>
Date: 2016-05-29T19:34:57Z
From: Achim Zeileis
Subject: zelig package: robust SE
In-Reply-To: <8c76c3349731445dba5ba8feca07daf1@ex13-live-mbn1.ad.kent.ac.uk>
On Sun, 29 May 2016, T.Riedle wrote:
> Dear R users,
>
> I am trying to run a logistic regression using zelig. The simple
> logistic regression works well but now I want to have HAC robust
> standard errors. I have read in the manual that there is an option
> called "robust" and that zelig() computes robust SE via the sandwich
> package. However, it doesn't work. My code looks as follows:
>
> crisis_bubble4<-zelig(stock.market.crash~crash.MA+bubble.MA+MP.MA+UTS.MA+UPR.MA+PPI.MA+RV.MA<http://rv.ma/>,robust=TRUE,model="logit",data=Data_logitregression_movingaverage)
>
> Error in glm.control(robust = TRUE) : unused argument (robust = TRUE)
Possibly this changed in recent versions of "Zelig". The documentation
for relogit on the official Zelig web page shows the same error:
http://docs.zeligproject.org/en/latest/zelig-relogit.html#example-2-one-tau-with-weighting-robust-standard-errors-and-bias-correction
In any case, Zelig only interfaced the so-called "robust" standard errors
(sandwich or HC0) and not the ones with HAC correction.
> I took this code from the zelig manual and don't understand why I get an
> error. What is more, I want to calculate NeweyWest SE or the SE using
> the weights by Andrews via the kernHAC function. How can I do that?
Why do you want to use Zelig? Apparently, the solution using plain glm()
along with the sandwich package and coeftest() did work for you. And in my
other e-mail I showed you how you can use lrm().
> Thanks for your support.
>
> Kind regards
>
> [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>