I ran Maureen's data. The correct p-value for a test should be zero
for X2, and for both D-s. The problem, which I fixed long time ago, is
provoked by the large number of alleles relative to the 2*N (total
number of genes), although I did not know about that effect of the bug
on p-values for the conditional X2. The version on
ftp://brooks.statgen.ncsu.edu/pub/zaykin/chihw.exe
is a correct one (it is a self-extracting archive). I could've sent a
few buggy copies to people before.
Below is a test sample (34 alleles) that should give a 0 probability
for the conditional X2, and probabilities smaller than 0.01 for both
D, and abs(D). A buggy one will report p-value for Cond_X2 around
90%. If you are using a buggy CHIHW, please ftp the correct one, or
email me and I will send you a copy.
ON:7 Is:3 Tj:6 Fu:8
ob:1 lI:1 jd:3 dq:7
Mi:2 UN:9 Jr:6 rT:1
kJ:3 Xe:5 ba:9 ry:9
fF:8 ue:2 lC:2 Vy:9
Jb:5 yL:8 nK:1 XT:5
cU:6 bF:3 Pw:2 yi:4
tX:9 dy:1 uY:1 lN:2
Dmitri
