Hi,
I was getting large variances for the heritability estimated by
ASREML (PVS file) so I decided to work through the equation in the manual
by hand to
understand it better. I also compared the results with Dickerson's
method cited in Dieters et al. (1995) T.A.G. 91:15-24. Dieters et al.
claimed that Dickerson's method gave similar results but slightly more
conservative (larger) than the Taylor method (used in ASREML).
The results were:
ASREML Var(herit) = 0.155 SE(herit) = 0.393
Hand calc Var(herit) = 0.024 SE(herit) = 0.155
Dickerson's Var(herit) = 0.032 SE(herit) = 0.179
Dickerson's method was more conservative than the hand calculation as
reported by Deiters et al but the ASREML output in the PVS file was much
larger and looks like the S.E. and not the variance as stated in the
manual.
Can someone help me with this?
The data I used for the hand calculations was extracted from the relevent
ASREML files and is included below for your information.
Variance components from the ASR file:
Source Model terms Gamma Component Compnt/StndErr
Trait.Family 225 69 0.246797 428.667 2.95
Trait.Site.F 450 130 0.147830 256.769 3.11
Variance 2108 1723 1.00000 1736.92 28.45
Phenotypic variance = 2422.356
>From VVP file:
Variance of Variance components
21082.6
-3608.44 6822.36
21.4726 -408.195 3726.72
Variance of Phenotypic variance = 23641.355
Herit = 3*FamVar / PhenVar (coefficient of relationship between full and
half-sib)
Thanks,
Alastair