does hill_sat go through (0, 0)

[dirknbr]

Is it possible that your hill_saturation function doesn't pass through (0, 0), which would be necessary since we assume 0 sales if a single channel is 0

[juanitorduz]

Thanks for reporting this @dirknbr ! We have a fix in #920

[dirknbr]

but does this fix mean that sigma is still the aymptote for all positive values of lambda and beta?

[juanitorduz]

but does this fix mean that sigma is still the aymptote for all positive values of lambda and beta?

Actually, it's going to be sigma minus the offset. Let me reopen this issue and see if we can bring back the original parametrization. Maybe @dirknbr you wanna give it a go? :)

[dirknbr]

why not use equation 4 of https://static.googleusercontent.com/media/research.google.com/en//pubs/archive/46001.pdf

[juanitorduz]

Ok! Seems the way to go. I am unsure about the current parameterization via exponentials.

[juanitorduz]

why not use equation 4 of https://static.googleusercontent.com/media/research.google.com/en//pubs/archive/46001.pdf

Ok! Sterted a PR #925

[juanitorduz]

We added the original implementation of the paper (with the corresponding reference)