1

When calculating the predicted probabilities in a logistic regression model, do we consider all the variables or just the significant ones? For eg: Let's say my model has: dependent variable Y and 3 dependent variables Xi out of which coefficients of X1 an X2 are significant whereas X3 is not significant. So for calculating the the predicted probabilities will I use just X1*beta1 + X2*beta2 or include X3*beta3 as well?

Punar
  • 13
  • 1
  • 3
  • possible duplicate of [Is it necessary to exclude all nonsignificant parameters to choose the best model?](http://stats.stackexchange.com/questions/90899/is-it-necessary-to-exclude-all-nonsignificant-parameters-to-choose-the-best-mode) – Scortchi - Reinstate Monica Nov 19 '14 at 10:12
  • Also [Should covariates that are not statistically significant be 'kept in' when creating a model?](http://stats.stackexchange.com/questions/66448/) – Scortchi - Reinstate Monica Nov 19 '14 at 10:17

1 Answers1

0

That question was asked and answered here: http://www.statalist.org/forums/forum/general-stata-discussion/general/450768-calculation-of-predicted-probabilities

Maarten Buis
  • 19,189
  • 29
  • 59