[R-sig-ME] Random or Fixed effects appropriate?

Nick Isaac njbisaac at googlemail.com
Wed Apr 9 19:47:15 CEST 2008

Thanks again for taking the time to reply.

>You were not clear on how you plan to use the conditional modes
(i.e., your point 1)
I have two y variables for which there is very little overlap in the
species that were observed. However, there is plenty of overlap at
coarser taxonomic resolution (random effects A, B, C and D). I was
interested in whether y1~x and y2~x vary among taxa A, B, C and D. So
I treated species as the units for analysis and the taxonomy as nested
random effects. To my surprise, the best model for y1 and y2 has an
identical structure: y ~ x + (1|A) + (1|B) + (x|C) + (1|D). So it was
then that I became interested in comparing x|C from the two models.

>Please keep in mind that conditional modes are not independent
"observations" like a >group mean or within-group effect or slope
This was at the back of my mind (poorly formulated) when I write that
using conditional models felt like cheating.

Best wishes, Nick

More information about the R-sig-mixed-models mailing list