LD h2 score substantially greater than 1

[Brad Verhulst]

Hi All, 

I am getting a h2 that is substantially larger than 1 (the output is pasted below). I understand that there is some wiggle room in the 0-1 boundary for estimating the heritability (and it is not significantly different than 1 - or 0 for that matter), but I am wondering if there is an obvious error that that I am making that is driving this result. It is also possible that I am interpreting this entirely wrong.

Any insights would be much appreciated.

Total Observed scale h2: 143.4505 (164.3891)

Lambda GC: 1.0165

Mean Chi^2: 1.0161

Intercept: 1.0073 (0.0076)

Ratio: 0.4535 (0.4736)

[Raymond Walters]

Hi Brad,

Two possibilities that come to mind:

1) Is this a binary phenotype with a very, very low prevalence? That can make the liability scale estimate wildly unstable (hence the giant SE accompanying your h2 estimate). Running the observed scale estimate could help check this. 

2) Is it possible you've misspecified the sample size? This result would be consistent with understating sample size by a couple orders of magnitude. The log from munge_sumstats should help verify how ldsc read sample size.

(If I had to guess, I'd start with the second option.)

Cheers,

Raymond

--
You received this message because you are subscribed to the Google Groups "ldsc_users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ldsc_users+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/ldsc_users/ac028573-4e40-4354-8b6e-03e521ed264b%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Brad Verhulst]

Thanks Raymond, 

I really appreciate your reply. It was certainly the second option.  Somehow, and I don't know how this happened, I specified sample size, e.g. using -- N 7516, but in the munge file it said sample size column is N.  In my case N was the number of samples in the meta analysis (3 in fact - so yes, perhaps a little small). When I changed the column name munge did not recognize the column as sample size, and then took the --N value instead. 

Thanks again, 

I really appreciate it.

Brad

On Thursday, September 21, 2017 at 10:35:24 AM UTC-4, Raymond Walters wrote:

Hi Brad,

Two possibilities that come to mind:

1) Is this a binary phenotype with a very, very low prevalence? That can make the liability scale estimate wildly unstable (hence the giant SE accompanying your h2 estimate). Running the observed scale estimate could help check this. 

2) Is it possible you've misspecified the sample size? This result would be consistent with understating sample size by a couple orders of magnitude. The log from munge_sumstats should help verify how ldsc read sample size.

(If I had to guess, I'd start with the second option.)

Cheers,

Raymond

On Sep 21, 2017 9:47 AM, "Brad Verhulst" <brad.v...@gmail.com> wrote:

Hi All, 

I am getting a h2 that is substantially larger than 1 (the output is pasted below). I understand that there is some wiggle room in the 0-1 boundary for estimating the heritability (and it is not significantly different than 1 - or 0 for that matter), but I am wondering if there is an obvious error that that I am making that is driving this result. It is also possible that I am interpreting this entirely wrong.

Any insights would be much appreciated.

Total Observed scale h2: 143.4505 (164.3891)

Lambda GC: 1.0165

Mean Chi^2: 1.0161

Intercept: 1.0073 (0.0076)

Ratio: 0.4535 (0.4736)

--
You received this message because you are subscribed to the Google Groups "ldsc_users" group.

To unsubscribe from this group and stop receiving emails from it, send an email to ldsc_users+...@googlegroups.com.

[Raymond Walters]

Hi Brad,

Glad that solved it! For future reference, if you see munge_sumstats.py misinterpreting a column like that in the future, it can often be solved without needing to edit the file by using the --ignore flag to list the names of columns that should be excluded.

Cheers,

Raymond

To view this discussion on the web visit https://groups.google.com/d/msgid/ldsc_users/51f27d0a-e680-4f8c-9e63-fcff737a3f2d%40googlegroups.com.