Add more statistics to the output table

[roninsightrx]

Implements #21 . The original request was r^2, but sum-of-squares makes more sense.
OFV and weighted sum of squares we need for model averaging (#27 ), on their own they are not so useful as a statistic.

Either OFV or SS can now easily be used to compute a weighted prediction (=model averaging).

[Copilot]

Pull Request Overview

This PR adds additional statistics (OFV and weighted sum-of-squares) to the output table to support model averaging and provide more comprehensive evaluation metrics.

• Introduces a new ss() function to calculate weighted sum-of-squares of residuals
• Adds ofv and ss_w columns to the results output table
• Updates tests to verify the new columns and their calculated values

Reviewed Changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
R/misc.R	Implements the new ss() function for weighted sum-of-squares calculation
R/run_eval_core.R	Integrates OFV and weighted sum-of-squares into the results table
man/ss.Rd	Documentation for the new ss() function
tests/testthat/test-run_eval.R	Test updates to verify new statistics columns and values

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[mccarthy-m-g]

Looks good! Just some minor suggestions:

[mccarthy-m-g]

Suggestion for something more concise (you can take or leave). vctrs::vec_recycle_common() will throw an informative error if any of the lengths are different, e.g.:

ss(obs = 1:5, pred = 5:1, w = 1:3)
#> Error in `ss()`:
#> ! Can't recycle `obs` (size 5) to match `w` (size 3).

Suggested change

	#' Weighted sum-of-squares of residuals
	#'
	#' @inheritParams rmse
	#' @param w weights
	#'
	ss <- function(obs, pred, w = NULL) {
	if(is.null(w)) {
	w <- rep(1, length(obs))
	}
	if (length(obs) != length(pred) || length(obs) != length(w)) {
	cli::cli_abort("`obs`, `pred`, and `w` must have the same length")
	}
	if(sum(w) == 0) return(NA)
	sum(w * (obs - pred)^2)
	}
	#' Weighted sum-of-squares of residuals
	#'
	#' @inheritParams rmse
	#' @param w A numeric vector of weights the same size as `obs` and `pred`.
	#' Vectors of size 1 will be recycled to the size of `obs` and `pred`.
	#'
	ss <- function(obs, pred, w = 1) {
	args <- vctrs::vec_recycle_common(obs = obs, pred = pred, w = w)
	if(sum(args[["w"]]) == 0) return(NA)
	sum(args[["w"]] * (args[["obs"]] - args[["pred"]])^2)
	}

Here are some tests you could add for this too:

test_that("recycling + incompatible size error works", {
  expect_equal(ss(obs = 1:3, pred = 3:1, w = 1), 8)
  expect_equal(ss(obs = 1:3, pred = 3:1, w = c(1, 1, 0.5)), 6)
  expect_error(
    ss(obs = 1:5, pred = 5:1, w = 1:3), class = "vctrs_error_incompatible_size"
  )
})

[roninsightrx]

Looks good! Just some minor suggestions:

in principle good suggestion, but leaving as-is for now for consistency (we should then also update the other stats functions, rmse etc to use the same approach)