[R-sig-eco] mrpp() / meandist() confusion

[Arne Erpenbach]

Dear list,

I'm looking through some old code and I'm confused.
I'd be glad if someone would help me out.

I've got paired-samples repeated measures design. I used a distance
matrix of a community and calculated MRPP statistics for two groups,
with stratified permutations by samplings. I then had a look at the mean
within group distances.

MRPP gives me a (significant) effect size of A=0.02, with an observed
delta of 0.79 and an expected delta of 0.81. Class means are given as
0.81 and 0.7766, respectively. My interpretation was, according to my
own notes, and that groups were significantly different (but the effect
size was rather small). I also noted down than the first group showed a
slightly higher within-group beta-diversity.

However, I did *not* note anything on the meandist() results.
The mean within-group dissimilarities are 0.785, and 0.796, respectively.

I noticed this when going through the code today, and think it would
mean that the beta-diversity in the first group is actually a tiny bit
lower than in the second. I then did look at each sampling seperately,
but the pattern persists: mrpp() shows a *larger* class mean, and
meandist() a *smaller* mean of within-group distances.

Now I'm confused by looking at my own results.

Can anybody help me out?
(Really sorry if I'm just being stupid here.)

Regards,
  arne