Hi,
I have 100,000 observations and my dependent variable is binary. I need to run logistic regression but my bad rate is too low(less than 1%). If i build my model based on such low bads, model could be unstable.
One of the ways could be biased sampling wherein i can keep all the bad accounts and keep x % of good accounts and increase the bad rate to say, 3%.
Now i will run logistic regression with increased bad rate. However, after running the regression, i will have to correct my score of this biased bad rate. Can somebody help me understand how will we correct the bad rate to 1%?
I have 100,000 observations and my dependent variable is binary. I need to run logistic regression but my bad rate is too low(less than 1%). If i build my model based on such low bads, model could be unstable.
One of the ways could be biased sampling wherein i can keep all the bad accounts and keep x % of good accounts and increase the bad rate to say, 3%.
Now i will run logistic regression with increased bad rate. However, after running the regression, i will have to correct my score of this biased bad rate. Can somebody help me understand how will we correct the bad rate to 1%?