DW stat for Panel Data

[RheaMaria]

Dear all,

I was just wondering if the Durbin Watson Statistic is a reliable indicator of the presence of autocorrelation in my fixed effects model with GLS cross-section weights and clustered robust-standard errors.

It is at 1.56 with an R squared of 0.38

When I add in an AR(1) term to my common coefficients, the DW stat rises to 1.82 and the R squared rises to 0.42. However the line graphs of my residuals do not change.

Would really appreciate some help. :D

Thanks
Rhea