dropped observations when running logit

[Li Jiang]

Hi everyone,

I'm running a logit which returns the following message:
note: 65.city != 0 predicts success perfectly

65.city dropped and 4 obs not used

I can't understand what these messages mean, specially the one where observations are lost. Can anyone shed some light? I get that observations in city 65 always seem to have the same 0/1 category, but I can't understand why these are not used.

best wishes,

Li

[Carlo Lazzaro]

Li:
the message you got tells you that there's no variation in the regressand due to -65.city-.
As a consequence, Stata drops -65.city- all discards all the observations related to it.

[William Lisowski]

Expanding on Carlo's answer, you can see a longer discussion of this characteristic of logistic regression in the discussion of Model Identification at Example 2 in the full documentation for the logit command found in the Stata Base Reference Manual PDF included in your Stata installation and accessible from Stata's Help menu.