# Alternative measures of prediction power?

What are alternative measures of the prediction power of a model, apart from the coefficient of determination $R^2$? What are their strengths and weaknesses, especially in comparison to the $R^2$?

-
What kind of model? Also, $R^2$ is not a measure of prediction accuracy, exactly. I suppose it could be thought of as a measure of within-sample prediction accuracy, but predictive power often refers to out-of-sample prediction accuracy. –  Macro May 10 '12 at 12:26
Well what are commonly used measures for the most popular model types? –  lodhb May 16 '12 at 7:43
One option is to withhold some data during the model fitting process and seeing how well the withheld data is predicted by the model (using, for example, mean squared error) and repeating this process for different randomly chosen subsets - see en.wikipedia.org/wiki/… –  Macro May 16 '12 at 11:11