0

I am performing research on Product attributes of mineral water toward purchase decision. However, the result of -2log likelihood test showed that my model is fit enough yet my result of Hosmer and Lemeshow is only 0.121. However, my Wald statistics showed none of my independent variables were statistically significant.

How should I interpret this result?

BGreene
  • 3,045
  • 4
  • 16
  • 33
christine
  • 1
  • 1
    This question is answered for multivariate regression at http://stats.stackexchange.com/questions/14500. The reasons given there apply equally well to logistic regression. – whuber Nov 19 '13 at 14:55
  • @whuber Is "multivariate" a slip for "multiple" (not that the key point doesn't generalise)? – Nick Cox Nov 19 '13 at 15:01
  • 1
    In addition to that, something else in particular to logistic models is that complete (essentially perfect prediction) or quasi-complete separation ("near"-perfect prediction, not sure how else to describe) can also result in a significant LRT but no significant variables (whose coefficients and SE will also be seriously inflated) – Affine Nov 19 '13 at 15:14
  • Thanks @Nick: that was indeed a slip. I meant *multiple regression,* not multivariate regression. – whuber Nov 19 '13 at 15:25

0 Answers0