01-12-2018 11:26 AM
I have built a predictive model in sas EM for predicting "no shows". The probability of high majority of cases of no show are between 0.57 to 0.68. Based on our business decision we came up with cut offs of 0.57. But I am not convinced to do overbooking with cases who have just crossed the cut off line like 0.59.
Isnot these cases have only 9 % more likely to be no show compared to random 50 50 chance ( like flip of coin)?
Thank you so much
01-13-2018 06:16 AM
IF you are using PROC LOGISTIC . Check options of MODEL.
CTABLE Displays the classification table
PPROB= Specifies probability cutpoints for classification