Ok, that's more helpful.
The problem is with replicate-weight designs, and it's because svyglm()
uses the fitted coefficients from the point estimate as starting values for
fitting the replicates. And even if that is changed, the computation of
the replicate variance doesn't like all the replicat
Well, I have uploaded the data in the public folder of my dropbox. Due
to data confidentiality, I haved to change the labels. To load the data:
con <- url( "http://dl.dropboxusercontent.com/u/101865137/datEx.rda"; )
print(load(con))
# The replicate weights were created according to the jackknife
On Fri, May 3, 2013 at 2:27 AM, Sebastian Weirich <
sebastian.weir...@iqb.hu-berlin.de> wrote:
> Hello,
>
> I want to specify a linear regression model in which the metric outcome is
> predicted by two factors and their interaction. glm() computes effects for
> each factor level and the levels of
Hello,
I want to specify a linear regression model in which the metric outcome
is predicted by two factors and their interaction. glm() computes
effects for each factor level and the levels of the interaction. In the
case of singularities glm() displays "NA" for the corresponding
coefficients
4 matches
Mail list logo
