[R-sig-ME] lmer: goodness of fit

[mikhail matz]

For lmer, is there a number, like an R-square, to add to the phrase in a paper "the model fit the data relatively well / relatively poorly"? Why is it wrong to simply report the correlation between predicted and observed values? 
I realize that it is a very frequently recurring discussion topic, so I would appreciate a simple direction to one of those previous discussions. Somehow I keep finding only the posts ask about percentage of variance explained by individual factors. What about the whole model? 

Misha