05-17-2015 06:44 AM
I am in the process of building a predictive logistic regression model. There is some seasonality in my data in the month of March-April. To mitigate seasonality effect, i was asked to take multiple sample windows and then merge them. My question - Should i include the March-April dataset while taking multiple sample windows? Or should i exclude that period? It has an impact on 1 variable. For example. Interest rate variable has changed mostly in seasonal period (march - april). My concern - the model may fail or less stable in out of time sample. How should i deal with it?