Variance-covaraince matrix V error

Hello,

I want to estimate the phenotypic variance explained by three sets of SNPs used for GWAS with eight traits. For one SNP set the analysis runs successfully for all traits. With the second SNP set the analysis runs well for six of the traits, but it gives the following error for the two remaining traits:

Error: the variance-covaraince matrix V is not positive definite

I did not find other posts on this forum mentioning this error. Is there a way to fix it?

Thanks,

af