Unfortunately, these standard errors aren't available for tobit model in
zelig.
Kosuke
--
Department of Politics
Princeton University
Hi Kosuke,
Sorry to write you directly with this issue, but I've been searching for
an answer this for a while. I found a similar question online, but not
the actual answer. If there's a known answer somewhere, just point me to
it and I'll figure it out:
Are you aware of any potential problem using robust SE and/or clustering
in Zelig's tobit models?
Thanks!
Below are some details, if necessary...
I'm using Zelig version 3.4-8 and R 2.10.1. I managed to get the exact
same regressions running using AER's tobit command, but not in zelig.
Since using Zelig would save me from having to program clarify-like
routines, I would very much like to get this working in Zelig.
The basic regression without clustering and robust SE's works fine:
tob1 <- zelig(mrelideal~relideo+incab, ,model="tobit",data=data.set)
But if I attempt the robust se's or clustering I get:
zelig(mrelideal~relideo+incab, robust =
TRUE,model="tobit",data=data.set)
Error in 1:nrow(mrelideal) : argument of length 0
tob1 <- zelig(mrelideal~relideo+incab, robust
= TRUE, cluster="pty",
model="tobit",data=data.set)
Error in rowsum.default(resid(fit, "dfbeta"), cluster) :
'x' must be numeric
I've tried converting pty to numeric, I got rid of rownames, I
experimented using the name of the data frame with the variables, but I
get the same errors. Any obvious error or known issue?
As I mentioned, with AER's tobit it works
tob1 <- tobit(mrelideal~relideo+incab,data=data.set)
robust.se <- coeftest(tob1, vcov=sandwich) #but not with vcovHC
clustered.se <- clx(x, 1, data.set$pty) #Where the clx is a function I
found online that is built on coeftest
-----------------------------------------------------
Cesar Zucco Jr.
Lecturer in Politics and International Affairs
Princeton University
www.princeton.edu/~zucco
-----------------------------------------------------
-
Zelig Mailing List, served by Harvard-MIT Data Center
Send messages: zelig(a)lists.gking.harvard.edu
[un]subscribe Options: