Time-series regression - shape error / Input dimension mis-match

A couple things that might help:

  • if alpha_zero is the starting point for the time varying params, you don’t want to pass it as mu to the Random Walk. Mu in the random walk is for drift - the mean of the errors.
  • you don’t need to tile step_size_ if it’s the same for all observations

If that doesn’t fix, I’d triple check that len(pred_cols) == sample_shape == 19

1 Like