Feature Selection with p-values

Feature selection through p-values using stats from ordinary least squares(OLS) is relevant if the data given does not satisfy pre requirements for linear regression ?

The issue is that a correlation within one data set is not necessarily causative. This is where domain knowledge is important as that is one of the key issues in feature selection. After all, these are dynamic systems, if all we needed were P-Values to pick features then we’d always have perfect forecasts which is not the case.

© Copyright 2013-2020 Analytics Vidhya