One question for the F-test

[Lisaliu]

Thanks for your help!

I was thinking like that in your first paragraph. So B1,B2…Bn are all compromised in representing the true value each time the regression model is taking an observation. Therefore these compromises add up leading to the overall compromise which is then the regression residual.

I forgot to differentiate between pos. and neg. serial corr… So if all the standard errors are correlated positively, that makes slope coefficients’ job easier for explaining the DV, cause the standard errors did that job to some extent, inflating the slope coefficients’ explaining power and underestimating their standard errors…

[tickersu]

@Lisaliu

Thanks for your help!

I forgot to differentiate between pos. and neg. serial corr… So if all the standard errors are correlated positively, that makes slope coefficients’ job easier for explaining the DV, cause the standard errors did that job to some extent, inflating the slope coefficients’ explaining power and underestimating their standard errors…

I would try not to think of it in that way. Serial correlation doesn’t “help” the independent variable explain the DV. Positive serial correlation can make an independent variable appear to be more statistically significant for predicting the DV, by way of inflating the t-test (since the standard error of the coefficient is underestimated).

[Lisaliu]

Thanks! You are right. And in Time-Series Analysis serial correlation is better explained using graphs and everything. Move on through the readings…!