[GenABEL-dev] faster polygenic

[Yurii Aulchenko]

>> 2) still, FMM generates L which are>1 (1.014), and this is VERY
>> significant. Old and hglm-based tend to be a bit on other side -- L<
>> 1 (0.996 for old, 0.993 for hglm), and for hglm this is more
>> pronounced
>
> That's interesting, I guess the extra variance inflation is probably due to
> uncertainty in K which we don't model. I've wondered about extending the
> method to work with an uncertain K, I'm not sure how feasible statistically
> or computationally that would be.

This is very interesting suggestion -- never quite thought of what
woul the effects of uncertainty of K be onto test statistics. My
feeling is that uncertainty in K is likely to translate into
uncertainty of Lambda (so, SD of L would be bigger), but not into bias
-- and that is what we see; but I can well be wrong about that.

I think we can get a better view of the problem and answer to
William's hypothesis when we get polygenic-FMM working as an entity,
separate from FMM-GWAS and generating standard 'null polygenic'
outputs. Now we kind of compare apples and pears by contrasting
old/hglm-null-polygenic + MMSCORE-GWAS with null-polygenic-FMM +
FMM-GWAS, and FMM-GWAS as a method is different from MMSCORE. So I
really wonder what would come out if we generate null-polygenic with
all three methods and can push the GWAS later on with both MMSCORE and
FMM. I believe we may have important clues on the above formulated
problem from that comparison.

best wishes,
Yurii