mmm.plot_posterior_predictive() does not work

[Garve]

Hi! I trained a model, and when I execute
mmm.plot_posterior_predictive(), I get

---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
Cell In[262], line 1
----> 1 mmm.plot_posterior_predictive()

File [~\miniconda3\envs\mmm\Lib\site-packages\pymc_marketing\mmm\base.py:303](http://localhost:8888/lab/tree/Desktop/MMM/~/miniconda3/envs/mmm/Lib/site-packages/pymc_marketing/mmm/base.py#line=302), in BaseMMM.plot_posterior_predictive(self, original_scale, **plt_kwargs)
    300 def plot_posterior_predictive(
    301     self, original_scale: bool = False, **plt_kwargs: Any
    302 ) -> plt.Figure:
--> 303     posterior_predictive_data: Dataset = self.posterior_predictive
    304     likelihood_hdi_94: DataArray = az.hdi(
    305         ary=posterior_predictive_data, hdi_prob=0.94
    306     )[self.output_var]
    307     likelihood_hdi_50: DataArray = az.hdi(
    308         ary=posterior_predictive_data, hdi_prob=0.50
    309     )[self.output_var]

File [~\miniconda3\envs\mmm\Lib\site-packages\pymc_marketing\mmm\base.py:251](http://localhost:8888/lab/tree/Desktop/MMM/~/miniconda3/envs/mmm/Lib/site-packages/pymc_marketing/mmm/base.py#line=250), in BaseMMM.posterior_predictive(self)
    248 @property
    249 def posterior_predictive(self) -> Dataset:
    250     if self.idata is None or "posterior_predictive" not in self.idata:
--> 251         raise RuntimeError("The model hasn't been fit yet, call .fit() first")
    252     return self.idata["posterior_predictive"]

RuntimeError: The model hasn't been fit yet, call .fit() first

The self.idata exists, but it does not have a posterior_predictive attribute.

pymc_marketing version 0.4.0.

[wd60622]

Did you set the extend = True when using the 'sample_posterior_predictive' method?

[Garve]

Hi!

I just trained a model and called this method. I didn't call sample_posterior_predictive explicitly anywhere.

Best
Robert

[smusch2]

@Garve Did you run the top line shown here? It fixed my issue.

mmm.sample_posterior_predictive(X, extend_idata=True, combined=True)
mmm.plot_posterior_predictive(original_scale=True)

[juanitorduz]

We need a better error message for this one.

[ulfaslak]

I have run into this too. How about adding some logic that does sample_posterior_predictive inside plot_posterior_predictive if it hasn't already been done.

[wd60622]

#741 introduces new error message telling user to call sample_posterior_predictive method