Skip to content
Prev 136954 / 398498 Next

Weird SEs with effect()

Dear John & Brian,

Ok. Now I start to understand. So basically  I cannot use the given SEs 
for my purposes. They only make sense on the scale of log-counts. 
However in a paper, the log-count scale is not very informative (the 
reader want to see the effect on the scale you measure). If I understand 
you right, the confidence intervals are fine though and maybe I will use 
them to illustrate the reliability of the estimate. The problem is that 
showing SEs of the adjusted means has become sort of the standard way to 
illustrate things in my field (too many SAS users ?) and I might run 
into trouble with the reviewers. I have several data sets with similar 
data and my plan was to use effect(). That is why I want to figure this 
out. I hope I haven't been too annoying ;).

Finally, is there a way to get correct SEs on the count scale (with 
adjusted means)??
I guess not, judging by your answers.

Thanks again for your help,

Gustaf
John Fox wrote:
-