Ok. I have conditional hetroskedacity down pat. For positive serial correlation: Do this mean that the presence of an error term on the positive side of the regression line will increase the chances of another (the next) error term on the positive side of the line? At first I thought each error term would get larger and larger, but this doesn’t make any sense.

Or does it mean that you have an increased chance of getting an error term on the same side of the regression line next time?

you got it. positive serial correlation means a positive error in time 1 increases the chance for a positive error at time t+1. Same with negative. This is first order serial correlation…which means one period to the next.

jb, We are talking about the error relative to the regression line aren’t we?

Yes…and it is very obvious when it is plotted on a scatterplot, as is heteroskedasticity. The thing to know about serial correlation, in addition to how to test it and fix it, is that assuming your model isnt an AR model, your parameter estimates will be okay. If one of the independent variables is a lag of the dependent, then you have a problem.

The scatterplot is kind of what confused me. It looked like an snaking S curve, with errors on both sides of the regression line. That is why I asked the second part of the question above: Or does it mean that you have an increased chance of getting an error term on the same side of the regression line next time. Is this what you were replying to when you said I got it?

mwvt9 Wrote: > Or does it mean that you have an increased chance > of getting an error term on the same side of the > regression line next time. more than 50%

Thanks.