0

The link test was ran for a possion model. enter image description here

This picture was taken out of the Negative binomial regression textbook by Joseph Hilbe.

Hessa
  • 3
  • 2
  • Welcome to CV. Please be aware that software specific questions such as yours are not encouraged. – Mike Hunter Jun 17 '16 at 22:46
  • Hilbe presumes his readers understand the material discussed at http://stats.stackexchange.com/questions/31/what-is-the-meaning-of-p-values-and-t-values-in-statistical-tests . – whuber Jun 18 '16 at 15:46

1 Answers1

2

If your model is properly specified, then you should not be able to find any additional predictors that are statistically significant. This command uses the predicted value from the poisson (_hat) and the predicted value squared (_hatsq) as the predictors and re-estimates the model. It's a kind of variable augmentation test.

The variable _hat should be a statistically significant predictor. This should be the case unless the model is utter junk. The variable _hatsq should be insignificant. If it is, this usually means that you omitted relevant variables and/or the link function is not correctly specified.

You can tell that you flunked by the low p-value or the confidence intervals on _hatsq not containing zero.

dimitriy
  • 31,081
  • 5
  • 63
  • 138