"In sample methods" for model selection often give conflicting results.Consider a dataset with 50 samples, and the following situations.

a) Suppose you were going to apply subset selection and you only have a few variables (less than 10), would AIC or BIC be a more appropriate method, and why?

b) Suppose you were going to apply subset selection had (p = 40), would AIC or BIC be a more appropriate method, and why?