Hello Freesurfer experts,
I am getting some
unusual results with the lme pipeline, and I would appreciate help with
identifying the most likely step in which the problem may have
originated.
Running lme_fit_FS.m on a single vertex, followed by
lme_F on the output of that (stats) yields a p-value of 0. This is because the fcdf function in matlab rounds to 1 if the F-value is big enough (in our case it rounds at about 80, and our F-value is 123).
This being the case, we are trying to determine whether this is (1) a problem with the data, (2) a problem with the script somewhere, or (3) a system setting I don't know about.
I'm not sure
if you have enough information, but can you tell me if the F-values and
Beta values look possible (reasonable?)?
Thanks for any advice you can provide.
-E
Starting Fisher scoring iterations
Likelihood at FS iteration 1 : 484.5267
Gradient norm: 53.7987
Likelihood at FS iteration 2 : 484.5984
Gradient norm: 1.3858
Likelihood at FS iteration 3 : 484.5985
Gradient norm: 0.022222
Likelihood at FS iteration 4 : 484.5985
Gradient norm: 0.00056884
Total elapsed time is 0.24038 seconds
stats =
Bhat: [7x1 double]
CovBhat: [7x7 double]
bihat: [1x93 double]
Covbihat: [93x1 double]
phisqhat: 0.0110
SIGMA: [345x5 double]
W: [345x5 double]
Dhat: 0.0213
X: [345x7 double]
Zcols: 1
re: [345x1 double]
ni: [93x1 double]
lreml: 484.5985
st =
1
K>> stats.Bhat
ans =
3.1531
-0.0559
0.0148
-0.0154
-0.0309
-0.0767
-0.0229
K>> -0.0559
K>> fstats = lme_F(stats,contrastmatrix)
fstats =
F: 123.4649
pval: 0
sgn: -1
df: [1 284.0811]