Checking for autocorrelation in residuals I have fitted a multiple linear regression equation and looked at the Durbin-Watson test statistic for the same.
I was told that in order to not rejecting null hypothesis of no auto-correlation against negative/postive auto-correlation, the p-value should be between 0.01 to 0.90.
Is above statement correct? Is there any reference available in internet to check the p-value in order to not-rejecting null?
Any pointer will be very helpful
 A: The Durbin-Watson statistical test can be used to test the absence of auto-correlation in a time series, it has a value in $(0,4)$, with values close to $2$ corresponding to time-series that do not have autocorrelation. Importantly though, while the D-W statistics can provide evidence for autocorrelation, it is not useful to determine the type of correlation. I would suggest looking at plots of the residuals (across time as well as against a lagged/shifted version of themselves) and the corresponding ACF/PACF.
Regarding the $p$-value you mentioned: I have not heard of this rule of thumb and it does seem like an arbitrary interpretation as we obviously need to utilise the value of the statistic itself too to determine if we have "negative/positive auto-correlation". For example, a $p$-value of $1.2*10^{-5}$ indicates that we reject at significance level $\alpha=0.05$ as well as $\alpha=0.01$ our null hypothesis $H_0$ of the correlation coefficient $\rho=0$, but it doesn't say what happens if $p$-value=$0.7$. The necessary $p$-value to reject our the null hypothesis $H_0$ is related to our significance level $\alpha$, if we decide that $\alpha=x$ and then any $p$-value below $x$ is "good enough" to reject the null.
PS. Avoid some "reference available in internet" for standard notions. Refer to a well-accepted reference on the subject (e.g. Wooldridge's "Introductory Econometrics", Rao's et al. "Linear Models
and Generalizations", etc.) at first instance. It will allow contextualising what is going on as well as serve as a reference point for additional specialised reads (often from the Internet).
