Thanks. We are aware of this problem and should be able to release a new
version very soon.
Kosuke
--
Department of Politics
Princeton University
http://imai.princeton.edu
On Fri, 2 Jan 2009, Jeroen Ooms wrote:
Happy new year Zelig people! :)
I am trying to use Zelig to simulate from a logit.gam model, however I
find that it gives me only expected values of 1, which is not correct
I think. The fitted model seems ok, but setx/sim somehow messes up the
scaling of the x variable. An example
x <- rnorm(1000)
y <- (x+rnorm(1000)) > 0
mydata <- data.frame(x=x,y=y)
z.out <- zelig(y~s(x),data=mydata,model="logit.gam")
x.out <- setx(z.out)
s.out <- sim(z.out,x.out)
summary(s.out)
#this is all nice what you would expect. Now let's rescale x.
mydata$x <- mydata$x + 10
z.out <- zelig(y~s(x),data=mydata,model="logit.gam")
x.out <- setx(z.out)
s.out <- sim(z.out,x.out)
summary(s.out)
this second model simulates only expected values around 1, although it
should be the same model as the first one, but with higher x values,
so if no x is specified in setx(), I would expect the same expected
values. Why does this happen?
I am using R 2.8.0, Zelig 3.4.0 and mgcv_1.4-1.1
Jeroen
-
Zelig Mailing List, served by Harvard-MIT Data Center
Send messages: zelig(a)lists.gking.harvard.edu
[un]subscribe Options:
http://lists.gking.harvard.edu/?info=zelig
Zelig program information:
http://gking.harvard.edu/zelig/
-
Zelig Mailing List, served by Harvard-MIT Data Center
Send messages: zelig(a)lists.gking.harvard.edu
[un]subscribe Options:
http://lists.gking.harvard.edu/?info=zelig
Zelig program information:
http://gking.harvard.edu/zelig/