Role of R-squared

[naruto]

Hi people, if i set my null hypothesis as h(o):rho=<0 ( no positive serial correlation) and i do a DW (durbin watson) test and get 0.5 which means there is misspec and high serial correlation, basically to reject h(o). Then i do a GLS using AR(1) and got a 1.70 for DW which after referred to the critical d values tells me to not reject the null h(o) and hence, conclude there is no positive serial correlation. Pls advise whether i have interpreted and conclude correctly -

1) The R-squared increased to 0.99 in AR(1) from 0.98 - does this indicates a better fit?

Thanks

[jiya]

yes you are correct....but notice that after correction your t-value of coefficients is very low while R-square is quite large, so there is an evidence of multicollinearity. Do a multicollinearity test of your origunal variables before you interpret your findings.