fbpx

 

Supplementary note for new function vpredict() in ASReml-R version 4

Jin Sun

 

vpredict() Compute functions of variance components and their approximate standard errors.

Functions of variance components and their standard errors can be obtained from the vpredict() function. As the variance parameter names can sometimes be long or unwieldy, the variance parameters are represented in vpredict() by the strings “V1”, “V2”,: : : in the order in which they appear in the vparameters component of the ASReml object.

By default for simple models ASReml fits on the ‘gamma’ scale, which means that the returned variance components are variance ratios (For more details please see the ASReml-R Reference Manual Version 4). Therefore, you have to be careful with vpredict as ASReml-R does not automatically convert variance ratios to variance components, or their standard errors. You need to fit the model on the component scale if you want to use vpredict and the s.e’s. To do this add: residual = ~idv(units). For Example:

head(harvey)

head(harvey.ped)

harvey.ainv <- ainverse(harvey.ped)

adg0.asr <- asreml(y3 ~ Line, random = ~vm(Calf, harvey.aiv), residual = ~idv(units), data = harvey)

> summary(adg0.asr)$varcomp

                                          component      std.error       z.ratio      bound      %ch

vm(Calf, harvey.ainv)                      499.5204       500.5139     0.9980149          P        0

units!units                                273.1669       410.0206     0.6662273          P        0

units!R                                      1.0000             NA           NA           F        0

> #  Heritability
> vpredict(adg0.asr, hA ~V1 / (V1 + V2))    
    Estimate        SE
hA 0.6464716 0.5658723