Nonnormality and Nonconstant Error Variance

As explained in Chapter 2, the standard linear regression model assumes that the errors are normally and independently distributed with 0 means and constant variance. The assumption that the errors all have 0 means is equivalent to assuming that the functional form of the model is correct, an assumption that I’ll address in the next chapter on nonlinearity. This chapter discusses diagnostics for nonnormal errors and nonconstant error variance, ...

  • Loading...
locked icon

Sign in to access this content

Get a 30 day FREE TRIAL

  • Watch videos from a variety of sources bringing classroom topics to life
  • Read modern, diverse business cases
  • Explore hundreds of books and reference titles