Hi everyone,
I have a dataset of plant drought tolerance values, and my
original goal was to partition variance in
these values among the nested random levels Biome/Study site.
However, my drought tolerance data is significantly non-normal,
and is best-fit by a gamma distribution, according to the R
package fitdistr. Since lme4 doesn't seem to be able to work
with gamma distributions (I got the
same error as described in this thread:
https://stat.ethz.ch/pipermail/r-sig-mixed-
models/2010q2/003786.html),
I wanted to use zelig to determine which level accounts for the
greatest variance in this trait.
I tried specifying the following model:
zelig(V37 ~ tag(1|Biomes/Site_Numbers), data = Seasonal_sites,
model = "gamma.mixed") -> mod1
But I get the following error message:
"In mer_finalize(ans) : singular convergence (7)"
And my estimated variances go to infinity:
0: 54.614133: 0.263523 2.62573
1: -1594.5787: 0.00000 18.5619
2: -1917.2900: 0.00000 25.9591
3: -2469.0981: 0.0137130 45.6228
4: -2936.0296: 0.00000 73.0933
5: -3428.4451: 0.00000 119.754
6: -3910.9637: 0.000840655 193.870
7: -4510.8977: 0.00000 352.310
8: -4903.5986: 0.00000 520.560
9: -5425.1380: 0.00000 873.876
10: -5896.6090: 6.68957e-09 1395.44
.
.
.
86: -42758.744: 8.50187e+09 1.05330e+19
87: -43234.416: 1.36270e+10 1.68826e+19
88: -43802.180: 2.39304e+10 2.96476e+19
89: -44250.090: 3.73145e+10 4.62293e+19
Why am I getting this odd result- am I doing something wrong?
The drought tolerance data "V37" varies
from roughly 1 -6, with n = 185, if that helps.
Please let me know if I can supply any more info about my data.
Thanks very much for your help!