[R-sig-ME] Significance of fixed effects. Kinship package (Marc Moragues)

Ben Bolker bbolker at gmail.com
Tue May 13 03:52:03 CEST 2014


Witold Proskura <witoldproskura at ...> writes:

> 
> Dear R Support Team,

  (Sadly there is no "R Support Team" -- just the people
who read the mailing list.)

> I have tried to compare the log-likelihoods of the two models, with and
> without the fixed effect which I want to test significance for. 
 
 [snip]

> Moreover please indicate some method for multiple comparisons of groups
> within particular fixed effect. When I tried to obtain df value for the
> model, the result was always NULL.
> The last question is how to extract variation components and to calculate
> heritability?

  We really need a reproducible example before we can help.
  What package is this from?  kinship (which is now obsolete),
or coxme?

  My only thought is that you have 'loglik' component that you
can probably extract via m1$loglik, and you can do something like

pchisq(-2*(diff. in log-likelihoods),df=(diff. in number of coefficients),
       lower.tail=FALSE)

to implement your own likelihood ratio test.


> My model:
> m1=lmekin(kgm~s1+rs+w+d+(1|id), data=milk,varlist=list(kmat))
> ,where:
> s1 and rs are categorical factors
> w and d are continuous variables
> 
> > summary(m1)
>              Length Class  Mode
> coefficients   2    -none- list
> var          576    -none- numeric
> vcoef          1    -none- list
> residuals    424    -none- numeric
> method         1    -none- character
> loglik         1    -none- numeric


 [snip]

> Linear mixed-effects kinship model fit by maximum likelihood
>   Data: milk
>   Log-likelihood = -3604.361
>   n= 424
> 
> Model:  kgm ~ s1 + rs + w + d + (1 | id)
> Fixed coefficients
>                    Value    Std Error     z       p
> (Intercept) -7920.701534 3023.8600619 -2.62 8.8e-03
> s1CT           76.794638  170.4547258  0.45 6.5e-01
> d              41.339062    9.9779674  4.14 3.4e-05
>

  [SNIP]

> Random effects
>  Group Variable Std Dev     Variance
>  id    Vmat.1      1210.785 1465999.783
> Residual error= 864.8264



More information about the R-sig-mixed-models mailing list