1

I am trying to perform autoregressive multiple linear regressions using statsmodels (something like y ~ y_1 + X1 + X2, not ARMA-like). More specifically, I'm looking for a way to get out of sample results. When I use the predict method, I get in sample results which means that it uses the previous historical value of the estimated variable instead of the estimated value of the variable.

Thanks for your help.

Tim Hutchison
  • 3,483
  • 9
  • 40
  • 76
Michael
  • 11
  • 2
  • `y_1 * X1 + X2` probably? – 9000 May 24 '17 at 14:34
  • Not really, to be more specific it has this (formal) form: y = constant + y_lagged1*a0 + X1*a1 + X2*a2 – Michael May 24 '17 at 15:13
  • There is no automatic support. If y_i is the lagged value, then you need to loop over each period and create the regressor matrix for predict for forecasting several periods. – Josef May 24 '17 at 20:45
  • OK, quite disappointing as it is very useful to assess predictive power of the model. Thansk for your feedback! – Michael May 26 '17 at 07:34

0 Answers0