deal with heteroscedasticity and autocorrelation in OLS

[koko_nothing]

hi.. i need some suggestion related to my thesis, it's about estimating equilibrium ex. rate (long run relationship)
i found all my variable stationer at level (10% level signfcn). when i run by OLS, it seem that my model got hetero & autocol problem..
i have 76 observations, 4 independent variables
i've tried ar(1) and newey-west HAC option, but this treatment make some variable no longer significant..
should I change my model (specification)? is there any appropriate method regarding the problems mentioned above?

thx in advance

[fmramos]

Hi I'm not so good with theory, but I'm used to make OLS estimation always with NW correction. It changes only your std. errors, not parameters. If u got not relevant variables after correcting your estimation, probably they're not relevant (or you can have a multicollinearity problem - I think this is really possible specially in a level regression - and, in this case, u can keep your variables without more serious problem or use pdl estimation). Good luck, Fabio