Apologies if this is too much scope-creep; feel free to close if it's not a good fit for this crate.
I've been playing with some data that's reported daily in some periods, but only monthly in other. Trying to fit the combined data then unsurprisingly focused more on the daily observations -- of which are are far more -- then the monthy ones. I've been working around that by just repeating out the monthly values to make fake daily ones, but that feels gratuitously inefficient.
It would thus be nice to be able to provide a weight for each data point as well. That would potentially enable approaches like https://en.wikipedia.org/wiki/Iteratively_reweighted_least_squares as well for other kinds of fits. (Repeating points works ok for integer weights, but is particularly annoying and slow for rational weights.)
Apologies if this is too much scope-creep; feel free to close if it's not a good fit for this crate.
I've been playing with some data that's reported daily in some periods, but only monthly in other. Trying to fit the combined data then unsurprisingly focused more on the daily observations -- of which are are far more -- then the monthy ones. I've been working around that by just repeating out the monthly values to make fake daily ones, but that feels gratuitously inefficient.
It would thus be nice to be able to provide a weight for each data point as well. That would potentially enable approaches like https://en.wikipedia.org/wiki/Iteratively_reweighted_least_squares as well for other kinds of fits. (Repeating points works ok for integer weights, but is particularly annoying and slow for rational weights.)