bayescan q values are generally higher than 0.05
0
0
Entering edit mode
12 weeks ago
Maxine • 0

I have 822K variants in 12 samples (which are separated into 2 groups) to detect selective adaption using bayescan. All parameters I choose were default except for pr_odds. I set pr_odds to 100.

The results I get are weird. The q values are pretty high (like 0.9) when p values are very small (less than 0.01). And the smallest q is only 0.27, which is higher than normally expected, right?

Here are the top 10 smallest p values results:

prop    log10(PO)   qval    alpha   fst
0.0036007   -2.442  0.98906 8.68E-04    0.11363
0.0038008   -2.4185 0.98906 -7.29E-04   0.11348
0.0038008   -2.4185 0.98906 -9.46E-04   0.11346
0.0040008   -2.3961 0.98906 -1.56E-03   0.11337
0.0040008   -2.3961 0.98906 -8.98E-04   0.11342
0.0040008   -2.3961 0.98906 -6.81E-04   0.11349
0.0042008   -2.3748 0.98906 -4.37E-04   0.11348
0.0042008   -2.3748 0.98906 1.44E-03    0.11368
0.0042008   -2.3748 0.98906 -6.05E-04   0.11349
0.0042008   -2.3748 0.98906 5.63E-04    0.1136


You can see the q values are very high.

Else, here are the top 10 smallest q value results,

prop    log10(PO)   qval    alpha   fst
0.72735 0.42613 0.27265 1.5232  0.3536
0.72414 0.41914 0.27425 1.5163  0.35244
0.72394 0.41871 0.27485 1.5188  0.35296
0.72374 0.41827 0.27521 1.4988  0.34917
0.72334 0.41741 0.2755  1.4995  0.34929
0.72234 0.41524 0.27586 1.4986  0.34924
0.71994 0.41005 0.27646 1.4917  0.34802
0.71874 0.40747 0.27706 1.5086  0.3515
0.71754 0.4049  0.27766 1.4945  0.34894
0.71734 0.40447 0.27856 1.4936  0.34871


No q value is less than 0.05, so as prop either.

Could anyone explain this situation? Many thanks.

statistics bayescan • 226 views
0
Entering edit mode

Is it possible that the sample size is too small (822k loci for multiple comparisons, but only 6 samples in a group)?