Dubin-watson statisc

usually how high signal a multicollinearity? thanks.

uh, wrong test. this tests for serial correlation

i thought DW was for Serial Correlation.

If it’s below 2 which is normally the lower limit, you’re screwed ie Auto-corr exists. DW has nothing to do with Multicollinearity…

Maybe my brain is fried, but do you mean serial correlation? DW=2.0 No serial correlation 0-2 positive serial correlation 2-4 negative serial correlation

Guys, it is more complicated than this Look up the two DW stats - DL and DU (Lower, Upper). Df = k, n 0 - DL = Positive Serial Corr DL - DU = Inconclusive DU - (4-DU) = Cannot Reject Null (4-DU) - (4-DL) = Inconclusive (4-DL) - 4 = Negative Serial Correlation Draw a little continuum and map it out.

CFAI doesnt even really talk about this. They just approximate it as 2(1-r) and if it is below the DW lower bound stat, then you have serial correlation.

For multicollinearity…if it is present you will normally see insiginficant t tests but significant F tests results. Also high R2. To correct it, you need to omit one of the variables that is causing the problem. i know i have alot of misspelled words. i am tired.