Unable to change 'method' for DelayedSaturatedMMM.plot_direct_contribution_curves() on pymc-marketing version 0.7.0

[ezjsiwu]

Hi there I'm rebuilding my MMM in pymc-marketing version 0.7.0 due to the newly available saturation and adstock capabilities. However, when I tried using the DelayedSaturatedMMM.plot_direct_contribution_curves() it gave me an error when I want to choose the method to 'michaelis-menten'..

[wd60622]

Hi there I'm rebuilding my MMM in pymc-marketing version 0.7.0 due to the newly available saturation and adstock capabilities. However, when I tried using the DelayedSaturatedMMM.plot_direct_contribution_curves() it gave me an error when I want to choose the method to 'michaelis-menten'..

Are you calling the method on the class itself? Or on an instance of the class?

Can you show the code you run leading up to this error message 🙏

[ezjsiwu]

hi Will, thanks for your response.. i'm calling on the method on the class itself..
Here's the code for my model config..

mmm = MMM(
    model_config = my_model_config,
    sampler_config = sampler_config,
    date_column="week",
    channel_columns=spend_cols,
    control_columns=control_cols,
    adstock="geometric", 
    saturation="logistic",
    adstock_max_lag=10,
    yearly_seasonality=10
)

mmm.build_model(X_train, y_train)


df_lift_test = pd.DataFrame.from_dict({
        "channel": ['spend_youtube','spend_linkedin','spend_facebook'],
        "x": [75900,48863.60,80000],
        "delta_x": [91000,10433,12000],
        "delta_y": [300,857,50],
        "sigma": [72,102,18]
    })

mmm.add_lift_test_measurements(df_lift_test)

mmm.fit(X=X_train, y=y_train, target_accept=0.90, chains=4, draws=1000, random_seed=123)

mmm.sample_posterior_predictive(X_train, extend_idata=True, combined=True)

Model was fitted fine and had acceptable results.. I just wasn't able to use the plot_direct_contribution_curve using method.. I ran this same exact code in version 0.6.0 and it worked fine..
mmm.plot_direct_contribution_curves(show_fit=True, method='sigmoid',xlim_max=3,same_axes=False)

[wd60622]

Model was fitted fine and had acceptable results.. I just wasn't able to use the plot_direct_contribution_curve using method.. I ran this same exact code in version 0.6.0 and it worked fine..

mmm.plot_direct_contribution_curves(show_fit=True, method='sigmoid',xlim_max=3,same_axes=False)

Try it without the method keyword then. It seems to have been dropped as the fit curve will come from saturation specified at initialization.

@juanitorduz shall we go in an a deprecation? Was this a keyword arg up until this release?

[juanitorduz]

Yes we can do this!

[ezjsiwu]

That works! Thanks @wd60622 :)