Contribution of each smooth term

[AdamWhiteXAI]

Is there anyway I can get the 'additional contribution' each smooth term makes to the final GAM score for a particular input? For example if my model is :
LinearGAM( terms=s(7) + s(13) + s(1) + s(15) + intercept)
then what contribution does s(7) make when 'variable 7' equals 10

Is it worth trying to hack into PyGam's code to get this?

[CodeCrux21]

what do you mean? it's the coefficient. it's an additive model.

[AdamWhiteXAI]

There isn't a single coefficient for a dependent variable in a GAM, it's value varies based on the value of the dependent variable.

[CodeCrux21]

There isn't a single coefficient for a dependent variable in a GAM, it's value varies based on the value of the dependent variable.

yes. and? it's contribution is still it's coefficient at x.

[AdamWhiteXAI]

But the point is that for a particular observation, you cannot get the single coefficient value that is being applied for to a specified independent variable. You can in R but not in PyGAM

[ouslan]

@AdamWhiteXAI What exactly is it that you want to implement? I imagine you are using the mgcv package from R. Could you link the documentation from R for the desired feature?