probit

[P. Hondelmann]

Dear List,

although looking in the archive I did not find the solution. I use the
Probit analysis in SPSS 12G. I have count data ("response", I use simply
the observed amount of dead individuals) vs dosis ("covariate"), the output
looked first ok. But the problem is: If I use subgroups for different
treatments (as "factor" 1 to n) the results differ completely compared to
the results of separate analyses. And a second problem: How to compare
different probabilites, i.e. LT50 from treatment 1 with LT50 from treatment
2? I have no idea since this are not means...
Thanks a lot for any input.

Peter

[Gary Rosin]

>I use the Probit analysis in SPSS 12G. I have count data ("response",
>...) vs dosis ("covariate"),.... But the problem is: If I use subgroups for
>different treatments (as "factor" 1 to n) the results differ completely
>compared to the results of separate analyses.

I haven't see a response to this question.  In the probit procedure,
the use of a factor only tests for a change in the intercept.  If you do
a split-file analysis, you can check for effects on the intercept *and*
on the coefficient(s).

Gary