Hi guys.
I have made two models, one let’s say exponential and one inverted s-shape. I have compared these two models with arviz compare and everything is fine and i can pass the weight list to weight_predictions.
df_comp_loo = az.compare({“exp_regression_model”: idata_exp,“s_regression_model”: idata_s})
idata_avg = az.weight_predictions([idata_s,idata_exp],weights=list(df_comp_loo.weight.values))
After checking out the Out of model predictions with PyMC - PyMC Labs i can make predictions on unseen data with either exp model or inverted s-shape model and works fine also. Is it possible to make predictions with the averaged model? I got stuck and I don’t know how to proceed after calling arviz weight_predictions.
For each model on its own I can do something like and obtain predictions.
new_x=[df.x.max()+1,df.x.max()+2]
with s_regression_model:
pm.set_data({“x”: new_x},coords={‘obs’:np.array(new_x)})
pp_s = pm.sample_posterior_predictive(idata_s,predictions=True, var_names=[“y”],random_seed=58)
Thanks a lot as always and if there is a link you could point me to i would appreciate it.