will high correlation between covariates in a regression model invalidate t-stat?