Skip to content
Prev 19183 / 20628 Next

Confidence interval around random effect variances in place of p-value

This would make an interesting simulation and/or theoretical exercise 
(I'm going to resist the urge to do it), i.e. identifying the 
correspondence between p-values constructed from parametric bootstrap 
full-vs-reduced model comparisons and p-values estimated as fraction of 
PB fits of full model that give variance=0 for the tested variance 
component(s).
On 4/2/21 8:22 PM, Jack Solomon wrote: