Hello,

I am running some analyses on the HCP, testing for the heritability of mean diffusivity along along various white-matter tracts. For one of these tracts I get a H2r of 0.985 and a warning message 'Unexpectedly high heritabilities might result from numerical problems, especially if mztwins are present.' This is the case even after multiplying by the suggested factor or when I transform the data using the inverse log. I don't have this issue for the other data and when plotting them side-by-side I cannot see an obvious difference. I am not sure what numerical problems may be causing this, if it is indeed an error. Any help would be much appreciated. My model looks like this:
solarPolygenic(formula = inv_log_MeanMD_M1S1 ~ Age_in_Yrs +
Age_in_Yrs^2 + sex + Age_in_Yrs * sex + Age_in_Yrs^2 * sex +
MeanFD, data = df, covtest = TRUE, screen = TRUE)
Best wishes,
Sam
p.s Thank you for the software, I use it all the time!

Hi Sam, Thank you for getting in touch.

The warning is simply triggered by high heritability value. It is to encourage you to do the sort of checking you have done.

High heritability should be taken in the context of the environmental factors including age and sex which have already been estimated and removed. Various other factors including unmodeled correlations may cause unexpectedly high heritability values, even 1.0 (not uncommon). Also don't forget to check the standard deviation of the heritability, as it might be difficult to estimate it very well with the present data. The most powerful transformation is inverse normalization such as:

define itrait1 = inormal_trait1
trait itrait1

I hope this helps. Please don't hesitate to contact us again for further clarification!

SOLAR-Eclipse Imaging Genetics Software Team