topic Re: Bias and Variance in Logistic Regression in SAS in Statistical Procedures

Bias and Variance in Logistic Regression in SAS

aranganayagi — Sat, 02 Nov 2019 17:54:57 GMT

HI, How do I check built Logistic regression model is suffering with BIAS or Variance.

I learnt about c-statistics, H-L Goodness -of-Fit Statistics, they all say how well the model is fit. But how can I interpret that the built model will perform better in unseen data. Please help to understand. Really struggling to get answer to this.

Re: Bias and Variance in Logistic Regression in SAS

aranganayagi — Sun, 03 Nov 2019 07:21:56 GMT

Can anyone help me on this

Re: Bias and Variance in Logistic Regression in SAS

Ksharp — Sun, 03 Nov 2019 11:57:10 GMT

I don't understand your Q well.

1) if you want know if model is robust ,

You can use train data and test data to check model . or

n-fold cross validate method( split your data into ten group, using 9 of 10 to build model and 1 of 10 to test, and again using another 9 of 10 to build model and 1 of 10 to test , over and over again for ten times).

2)if you want know if data is over disperse. try

proc logistic......

model ..../scale=none aggregate ;

3) if your train data is sample from a big table . you need

model ............./ offset=

to adjust your predict probability .