Multi Regression Assumptions and Violatons

Help! I understand multiple regression calculations and concepts of SS 12; although, I am having difficulty with the violations of the general assumptions: Hetero, Serial Corr, Multicollin. I do know the following (based on memorization): Hetero: residual not consistent BP=n*R^2 = standard error to compare Chi^2 critical value Serial Corr DW approximate 2(1-r) Mainly time series data Multi High correlation amoung independent variables look for high R^2 and F stat that is significant, but no T stats that are significant Can someone please clarify the general assumptions and the major violations. Thank you!

This is one of those things that you need to go back to your book and look at pictures. I would be happy to answer more specific questions, but your question here is pretty much a 50-minute class session with blackboard (every stat prof in the world has class notes for the "Multi Regression Assumptions and Violatons " lecture).

You might find this helpful: http://www.statsoft.com/textbook/stathome.html http://www.statsoft.com/textbook/stmulreg.html#assumptions

I am still working to grasp this but this is what is in my head, for now. Hetero: residual not consistent - think graphically. if for indep variables X from 0 to 10 the error term is small but for X from 11 to 20 the error term is large then the residual is not consistent and this assumption is violated. 20 seconds with a napkin and a pen will clarify this better than words. Serial Corr: -think coin flips and football. If you flip a coin twice and the first time it is heads the second flip is still just as likely to be heads or tails. f+1 is not correlated with f. If a football team wins game one at home by 28 then they are more likely to win game two at home. g+1 is correlated with g and the assumption is violated. Multi: -think pocket change. If Y is how much money is in your pocket and X1 is coins and X2 is quarters then X1 and X2 are correlated and the assumption is violated. I am not sure if this is dead on or if it helps clarify but it is a starting point for me.

Thanks!

Darien Hacker, great textbook! Thanks!

Can only underline the author who said you have to approach this with graphics. In text it looks quite complicated and with the graphic in front of you it’ll all make sense.