Hi,
I am fairly new to R and have encountered an issue with the lmRob function that
I have been unable to resolve. I am trying to run a robust regression using the
lmRob function which runs successfully, but the results are rather strange.
I'm not sure it's important, but my model has 3 dichotomous categorical
variables and 2 continuous variables in it. When I look at a summary of my
model, 1 of my continuous variables and 1 of my categorical variables both have
an "Estimate" value of 1. My other 2 categorical variables and 1 other
continuous variables all have "Estimate" values of 0. These results
seem strange to me. When I run a simple rlm function the model runs fine and
produces the results I'd expect to see.
Another potentially useful piece of information is that when I run a model with
just the two variables that have an "Estimate" value of 1 (one
categorical and one continuous), I get the following error message:
1: In lmRob.fit.compute(x, y, x1.idx = x1.idx, nrep = nrep, robust.control =
robust.control, :
Sum(psi.weight(wi)) less than 1e-06 in lmRob.ucovcoef. during initial
estimation.
2: In lmRob.fit.compute(x, y, x1.idx = x1.idx, nrep = nrep, robust.control =
robust.control, :
Sum(psi.weight(wi)) less than 1e-06 in lmRob.ucovcoef. during final scale
estimation.
I do not understand what this error message means and have been unable to
resolve it. I tried several different permutations of my five variables in
varying combinations and sometimes the models work and sometimes I get this
error message. There seems to be no pattern as to when the error message occurs
and when it does not.
If anyone has encountered this or has any help I'd appreciate it. Thanks.
[[alternative HTML version deleted]]
