For the delta deviance influence diagnostic in logistic regression ddeviance from predict (using v 14 right now, but I don’t think that this has changed), it seems that Stata uses formula #5.26 in Hosmer, et.al. (2013). Formula 5.26 is stated to be an "approximation" if the subject's Pearson residual is substituted for the deviance residual from the prior stated formula. However, Minitab uses the formula without this substitution (just prior, but unnumbered in the Hosmer text). These formulas produce somewhat different values, and I am trying to understand the logic of Stata on this choice.
Thanks for any enlightenment you might offer!
John
0 Response to ddeviance result differs from Minitab
Post a Comment