Document (saturation) transformations using a minimum number of plots

[cluhmann]

The adstock functions are documented using several different parameter values all plotted on a single chart (e.g., here). Some of the saturation functions are similar (e.g., the inverse scaled logistic, the tanh, and root)

But several of them use a parameterization-per-plot scheme (or one parameter varied in one plot and another parameter varied in another plot) which both more space-consuming and less useful (because you cannot compare the different curves as easily). Can we homogenize these? I like the way they are done in the PyMC documentation (e.g., here) with one plot and several example parameterizations.

Not sure this is all of them, but this is what I found that has the parameterization-per-plot scheme:

https://www.pymc-marketing.io/en/stable/api/generated/pymc_marketing.mmm.transformers.hill_function.html#pymc_marketing.mmm.transformers.hill_function

https://www.pymc-marketing.io/en/stable/api/generated/pymc_marketing.mmm.transformers.hill_saturation_sigmoid.html

https://www.pymc-marketing.io/en/stable/api/generated/pymc_marketing.mmm.transformers.michaelis_menten.html

[wd60622]

The title confuses me a bit. The goal here would be to vary only one parameter at a time and control all the others in order to better gauge the effect of each parameter on the curve, right? So the ask is to increase the number of plots?

In short, more or less plots? :smile:

[cluhmann]

In short, more or less plots? 😄

Fewer. I am basically suggesting one plot per transformation/function with that single plot illustrating multiple curves/functions/parameterizations/transformations. That's how the PyMC documentation works for individual distributions. But some of the PyMC-Marketing docs have several different plots/charts for a single transformation, each with a single parameterization (or sometimes several plots/charts with a couple per plot/chart). It's just very cluttered as it is currently.