1
> ncvTest(alm)
Non-constant Variance Score Test 
Variance formula: ~ fitted.values 
Chisquare = 121.2316    Df = 1     p = 3.400245e-28 

> spreadLevelPlot(alm)
Suggested power transformation:  4.428269

I am having issues adjusting my regression formula based on what the results of the non-constant variable test shows. How do I implement the suggested power transformation here? Obviously with such a low p value this is heteroscedastic. I have run a robust standard error linear regression already and it does not change the BP test p value. I have also performed a variance inflation factor test to see if multicolinearity is an issue here, which it does not seem to be.

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
confizzled
  • 11
  • 2
  • 4
    The main point of the spread-vs-level plot is to *look* at it, because what you see provides information well beyond a single statistic (the suggested power transformation). What does yours look like? (`R` code to construct this plot appears in an answer at http://stats.stackexchange.com/a/74594/919, which explains how to interpret it.) – whuber Dec 15 '13 at 19:16

0 Answers0