07-21-2011 04:41 PM
probably a naive question.
I'm estimating the variance component of the random factor of my Linear mixed model.
I get rather small component (of the order of 0.0002). This means that the random factor contributes little to the model.
But there is a way to derive how much variance is explained (that is a certain percentage) by the random factor by looking to the magnitude of the variance component of the random factor?
08-02-2011 10:32 AM
Whether or not 0.0002 is small or large would depend on the scale of your response variable. Regarding % variance explained, you might be able to do this, but it is not clear if you should do it. It depends on your model (and experimental layout). If you have a hierachcial variance component model, you could add all the variances and look at percentages. But note: the theoretical random effects in the model may be independent, but the estimates likely are correlated. The variance estimates would be correlated also. This complicates the interpretation in terms of percentages of a total.