Given multiple data sets (variables), I need to evaluate which set is the best predictor for my dependent variable (i.e., whether or not population affects annual sales, for example).

I ran a regression on each variable compared to the dependent variable - now which answer provides me with the "best predictor"? On one, for example, I received an r2 of 26.95%, but my t-statistic was -5.154. How can I have a t-stat which says there is no linear relationship, but an r2 that is telling me there is a high correlation (26.95% was the highest r2 received)?

Once I determine best and worst predictors, I need to determine what the best two-factor model is; how would I go about that? With 6 different variables to evaluate, it would take forever to evaluate each two-factor possibilities.

thanks!!