Skip to content
Prev 166736 / 398502 Next

Determining variance components of classed covariates

You might want to try the
https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
list next time for mixed model questions.

At any rate the Variance column figures are variances, not percentages.

We can use anova with REML=FALSE to make comparisons among
models.  Below we find that removing the rs7074431 term makes very
little difference so we can drop it but removing the rs11834524 term
makes a big difference.  Thus modx0 can be used.
Data: input_new
Models:
modx0: Expression ~ 1 + (1 | rs11834524)
modxx: Expression ~ 1 + (1 | rs11834524) + (1 | rs7074431)
      Df     AIC     BIC  logLik  Chisq Chi Df Pr(>Chisq)
modx0  3 111.386 119.460 -52.693
modxx  4 110.288 121.053 -51.144 3.0986      1    0.07836 .
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Data: input_new
Models:
mod0x: Expression ~ 1 + (1 | rs7074431)
modxx: Expression ~ 1 + (1 | rs11834524) + (1 | rs7074431)
      Df      AIC      BIC   logLik  Chisq Chi Df Pr(>Chisq)
mod0x  3  206.652  214.726 -100.326
modxx  4  110.288  121.053  -51.144 98.365      1  < 2.2e-16 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
On Mon, Jan 12, 2009 at 11:36 AM, Stephen Montgomery <sm8 at sanger.ac.uk> wrote: