Dear list members,
maybe this is too trivial - in that case, please accept my apologies.
If I perform two constrained ordinations (redundancy analysis in the vegan
package):
model1 <- rda(community_data ~ predictorA * predictorB)
and
model2 <- rda(community_data ~ predictorA:predictorB + Condition(predictorA
+ predictorB))
I would expect
anova(model1, by="margin")
and
anova(model2)
to be roughly equivalent, but the p-values tend to differ a lot, although
pseudo-F values do not. Why?
(Both permutation tests are run under reduced model, so just the residuals
are permuted. In my case, there is a simple block structure in the data, so
I do the permutation within blocks in both cases, but I do not see how this
should affect the results.)
Any advice, please?
With kind regards,
Martin Weiser
[[alternative HTML version deleted]]
_______________________________________________
R-sig-ecology mailing list
[email protected]
https://stat.ethz.ch/mailman/listinfo/r-sig-ecology
