Out of sample in cross-sectional model
Posted: Tue Apr 05, 2022 9:57 pm
Hello !
Im working with cross-sectional data fitting a Least Squares model.
n = around 400 obs.
In the past i have made a lot of forecasting exercises with time-series models, mainly dynamic forecasting technique through mobile window sizes. But in this case, i dont know which is a robust methodology for selecting the out of sample data in order to make a forecasting exercise with a cross-sectional model. My intuition tells me the diversity of the sample should be high, but its only a basic idea.
Can anyone suggest me a robust methodology? If there is some literature available, i would appreciate it a lot too
Thanks
Im working with cross-sectional data fitting a Least Squares model.
n = around 400 obs.
In the past i have made a lot of forecasting exercises with time-series models, mainly dynamic forecasting technique through mobile window sizes. But in this case, i dont know which is a robust methodology for selecting the out of sample data in order to make a forecasting exercise with a cross-sectional model. My intuition tells me the diversity of the sample should be high, but its only a basic idea.
Can anyone suggest me a robust methodology? If there is some literature available, i would appreciate it a lot too
Thanks