Error for heritability calculation using too small variances

[efukatsu]

Hi,

When I executed remlf90(), I obtained following results:

Variance components: Estimated variances S.E. genetic 1.844e-05 0.006452 Residual 3.721e+02 27.814000

           Estimate      S.E.

Heritability -9.729e-08 1.747e-05

Heritability should be 1.844e-05/(1.844e-05 + 3.721e+02) = 4.955657e-08

I guess the calculation of variance may be conducted using blupf90 with 16bit Half-precision floating-point format, so the variances smaller than 6.10x10^-5 may result in minus value. Is it right?

I would like to obtain advices about above results.

Thanks,

Eitaro

[famuvie]

Dear Eitaro,

Your estimate of the genetic variance is essentially 0 for all practical purposes. All variation is attributed to residual factors.

The negative heritability is just a numerical artefact of the genetic variance near the boundary of the parameter space and possibly some assumptions that are simply not met there.

ƒacu.-

On 15/09/2022 02:59, efukatsu wrote:

Hi,

When I executed remlf90(), I obtained following results:

Variance components: Estimated variances S.E. genetic 1.844e-05 0.006452 Residual 3.721e+02 27.814000

|Estimate S.E. |

Heritability -9.729e-08 1.747e-05

Heritability should be 1.844e-05/(1.844e-05 + 3.721e+02) = 4.955657e-08

I guess the calculation of variance may be conducted using blupf90 with 16bit Half-precision floating-point format, so the variances smaller than 6.10x10^-5 may result in minus value. Is it right?

I would like to obtain advices about above results.

Thanks,

Eitaro

— Reply to this email directly, view it on GitHub https://github.com/famuvie/breedR/issues/113, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABLOEDUBWYX3ZAUNJ7DL56LV6JYFZANCNFSM6AAAAAAQM5HE2I. You are receiving this because you are subscribed to this thread.Message ID: @.***>