Re: [Gretl-users] Threshold in Logit/Probit Models
Am 07.04.2019 um 17:17 schrieb Periklis Gogas:
Thanks a lot! I will try that!
On Sun, Apr 7, 2019 at 4:19 PM Allin Cottrell <cottrell(a)wfu.edu
<mailto:cottrell@wfu.edu>> wrote:
On Sun, 7 Apr 2019, Periklis Gogas wrote:
> Hello all!
>
> Can we set the threshold to alternative values than 0.50 in Logit
models?
> If not, it would have been a nice addition if it is easy.
> One step (or many more complex) would be to produce results for
i.e. 10
> alternative levels and also the corresponding ROC curve.
Try Peter Summers' "roc" function package. If you're unfamiliar
with
such packages see /Help/Packages.
Maybe this is redundant, but let me just add that the $yhat accessor
after logit contains the predicted probability, not a binary 0-1
outcome. So a user may map these probabilities to 0-1 with any threshold
she chooses.
(For example like: series y01 = ($yhat > 0.3))
cheers
sven