Quant methods #291
Quant methods #291
Conversation
|
Hey @dajmcdon. We'll take a look at this in about 2 weeks. Still in the "open heart surgery" changes to tune and company, trying to make a conference deadline. |
|
What Max said, i have 2 small things to work on in the meantime:
|
|
Thanks both! And no hurry. We'll probably want to handle #282 simultaneously. (FYI @brookslogan is a member of our team). |
EmilHvitfeldt
left a comment
EmilHvitfeldt
left a comment
There was a problem hiding this comment.
Lots of little comments. Overall looking good
R/impute-quantile_pred.R
Outdated
| if (any(indm)) { | ||
| qvals_out[indm] <- switch( | ||
| method, | ||
| linear = stats::approx(tau, qvals, tau_out[indm])$y, | ||
| cubic = Q(tau_out[indm]) | ||
| ) | ||
| } |
There was a problem hiding this comment.
If i'm not mistaken. Then I think we could avoid this switch() statement by moving it inside the if (middle == "cubic") { and if (middle == "linear" || any(is.na(result))) { correct?
R/impute-quantile_pred.R
Outdated
| indl <- tau_out < min(tau) | ||
| indr <- tau_out > max(tau) | ||
| indm <- !indl & !indr |
There was a problem hiding this comment.
Could you enlighten me what indl, indr, and indm indicates in this context? thank you!
There was a problem hiding this comment.
I've renamed these to be clearer. These are indexes that determine if the probabilities are being imputed (indm) versus extrapolated (the other two, outside the range of the existing probabilities).
Co-authored-by: Emil Hvitfeldt <[email protected]>
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
| }) | ||
|
|
||
|
|
||
|
|
There was a problem hiding this comment.
[air] reported by reviewdog 🐶
3 participants
This is the first of a few possible PRs that our team discussed with @topepo and @EmilHvitfeldt .
Features and bug fixes:
quantile_predobject.vec_ptype2andvec_castmethods forquantile_pred. Without these, occasional downstream bugs happened on the object.quantile_predobject, say the median. This way, rather than trying to "match" doubles, it can just compute it directly. It can also be used to expand the set of quantiles represented by the object.snap()andis_quantile_pred().