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

Which version of PyMC are you using? I suggest using v4 if you are not yet.

init, which is now called init_dist requires an unnamed distribution as input so

alpha_zero = pm.Normal.dist(mu=0,sigma=np.sqrt(2), shape=(sample_shape,1))

The new version would have given you a more informative error to explain this.

PS: I haven’t checked the rest of you code.