are these two the same thing? I thought autocorrelation means a time series correlates with its past lags of time series, while serial correlation means error terms are correlated. But I saw them used interchangeably, are these the same?
Yes, they are the same. They both describe the condition where a previous error predicts the next error. Note that there are two separate tests for serial correlation depending on the type of regression Linear Regression: Durbin Watson Autoregressive Model: rho(k)=((COVAR(x(t),x(t-1))/VAR(x)) , with a std error of 1/SQRT(T) Autoregressive serial correlation test is on page 454 of CFAI if the above doesn’t make sense.
thank you, ftwcfa