Skip to contents

Fit Prior Ridge

Source: R/prior_ridge.R
fit_pridge.Rd

Given a data.frame of matches, fit the pRidge model

Usage

fit_pridge(
  matches,
  priors,
  response_name = "score",
  grid = exp(seq(log(0.01), log(20), length.out = 100)),
  n_cores = NULL,
  digits = 2
)

Arguments

matches

(data.frame) dataframe of matches; assumed to have `red1`, `red2`, `red3`, `blue1`, etc. as team entries, and 2 columns representing the response named "(red/blue)_(responseName)"

priors

(numeric) vector of priors to regularize towards, named with tba-legal team identifiers (i.e. "frc449")

response_name

(character) name of the response vectors as they appear in `matches`

grid

(numeric) Grid of lambda values to consider for regularization

n_cores

the number of cores on your machine to reserve for calculation. If NULL, will default to the max - 1.

digits

(integer) number of digits to round the result to

Details

Selects lambda via LOOCV. Trims the priors to discard teams that don't appear in `matches`.

Examples

event_key <- "2026mdsev"
matches <- event_matches(event_key, match_type = "qual")
sb_data <- team_events_sb(event = event_key)
epas <- sapply(sb_data, function(te){te$epa$stats$start})
names(epas) <- sapply(sb_data, function(te){te$team})
mdsev_pridge <- fit_pridge(matches, epas)