In R, there are two implementations of the Breusch-Pagan test: bptest from the lmtest package and ncvTest from the car package. Besides the option of studentizing the result, the two functions differ in a critical way:

  • bptest fits residuals to model terms by default (i.e. X1, X2,...Xn)
  • ncvTest fits residuals to fitted values by default (i.e. ŷ)
  • Unfortunatley for my data (average DTI indices for several white matter tracts), the two options, when non-studentized, disagree quite often.

    If the two approaches differ substantially in their results, what would be an appropriate strategy to choose, and potentially justify, either approach?

    More Adam James Schadler's questions See All
    Similar questions and discussions