--het and --ibc Plink 1.9

[Katherine Daniela Arias Huamani]

Hi,

In the help, I found that Fhat2 should be approximately equal to the --het estimate. However, my results didn't show that:

with --ibc:

FID    IID    NOMISS    Fhat1    Fhat2    Fhat3
0    20    371924    -0.370288    -0.00840625    -0.189347
0    22    371859    -0.398575    0.178544    -0.110016
0    145    371620    -0.494473    0.0627794    -0.215847
0    66    372261    -0.504631    0.154813    -0.174909

with --het:

FID  IID       O(HOM)       E(HOM)        N(NM)            F
   0   20       220400    2.432e+05       371924      -0.1768
   0   22       244945    2.431e+05       371859      0.01417
   0  145       222900     2.43e+05       371620      -0.1558
   0   66       233887    2.434e+05       372261      -0.0735

The help also says that --read-freq could be used to correct small samples. I use --read-freq, however, the results are different too. 

¿Which one is correct? ¿Fhat2 or Fhet?

Thanks in advance.

Katherine 

[Christopher Chang]

These are estimators.  They are all likely to catch an actual instance of inbreeding, but they don't have identical behavior.  Only --het is implemented in Plink 2.0, but I wouldn't claim the --ibc estimators are "wrong".

Yes, the correspondence between Fhat2 and --het F is not as close as one would expect from the current language in the docs; I will edit the docs.