For my panel data (all continuous variables; 1 DV, 4 IVs; 8 time points for each of the 5 companies), the Breusch-Pagan test is nonsignificant (Chi-square(4) = 1.77291, p= 0.777434) implying that both Random effects model and fixed effect model are not appropriate.  I used the OLS regression with heteroscedasticy problem.  Any suggestions on how to remedy this?