Hello,
I'm looking for measure of goodness of fit of binary predictions made
by a statistical model. One such measure is Matthews correlation
criterion (MCC)[1]. However, MCC doesn't account for the number of
parameters used to create the model, which is crucial in my case.
Akaike Information Criterion (AIC)[2], does account for goodness of
fit and for the number of parameters that were used for model
creation. AIC involves calculation of maximized value of likelihood
function for that model (L). But I can't figure out how to calculate L
for the binary prediction case. Can anyone help me with this?
[1] http://en.wikipedia.org/wiki/Matthews_Correlation_Coefficient
[2] http://en.wikipedia.org/wiki/Akaike_information_criterion
|