Hi!
I'm trying to understand and reproduce some of your results, but I am having some issues with your implementation of the local fréchet weights for any problem (since they are all similar across repo) with more than one predictor.
Specifically, the formula you use in your code is quite different from the definition of s_{in} above equation (4.5) in Petersen & Müller (2019)
It's to be expected since your code handles p > 1, but when I use your formula with p = 1, the results do not coincide. In fact, this also shows when doing some numerical experiments that the weights obtained with your formula seem to give rather poor results.
I might be misunderstanding something here, so please let me know if I missed anything.
Best,
Matt
Hi!
I'm trying to understand and reproduce some of your results, but I am having some issues with your implementation of the local fréchet weights for any problem (since they are all similar across repo) with more than one predictor.
Specifically, the formula you use in your code is quite different from the definition of s_{in} above equation (4.5) in Petersen & Müller (2019)
It's to be expected since your code handles
p > 1, but when I use your formula withp = 1, the results do not coincide. In fact, this also shows when doing some numerical experiments that the weights obtained with your formula seem to give rather poor results.I might be misunderstanding something here, so please let me know if I missed anything.
Best,
Matt