2

In follow up to a question posted here: Constrained linear regression through a specified point

What are some ways that we can tell if a simple linear regression model is "good" when we constrain it through a specified point?

Community
  • 1
Mark Jones Jr.
  • 131
  • 3
  • What about "classical" testing, i.e. as a ratio of explained / unexplained variance and using F test? – Tomas Aug 13 '11 at 05:47

1 Answers1

5

Since forcing a regression line through a certain point means imposing constraints on the coefficients, you can do an $F$-test (R: anova) to compare the constrained and unconstrained models. And since this is still a regression model, the traditional stuff like residual diagnostics should make sense, at least partially: the residuals would not sum up to zero, for instance.

StasK
  • 29,235
  • 2
  • 80
  • 165