gvselect IC vs fitstat IC

[Annie Lalancette]

Hello, I ran a logistic regression using the ologit command and then calculated the information criteria using fitstat, ic. However, the values I obtain for AIC and BIC differ from those reported for the same models when running the command gvselect for best subset selection. Can anyone tell me why these values differ? For example:

ologit Rank ExpHaz Land Population
gvselect <term> ExpHaz Land Population: ologit Rank <term>

VS

ologit Rank Land
fitstat, ic

Assuming that Land is the best one-variable model that came out of gvselect.
Thank you!
Annie