sample_by_choice: Draw a choice-based sample stratified by the chosen...

In choicer: Discrete Choice Models for Economic Applications

Draw a choice-based sample stratified by the chosen alternative

Description

Subsamples whole choice situations from a population data set according to fixed per-stratum quotas, where strata are defined by the chosen alternative. The input data set is treated as the population, so the population shares Q(j) are known exactly; the returned sample carries a ready-to-use WESML weight column (see wesml_weights).

Usage

sample_by_choice(
  data,
  id_col,
  alt_col,
  choice_col,
  n_per_alt = NULL,
  frac_per_alt = NULL,
  seed = NULL,
  weight_name = ".wesml_weight",
  outside_opt_label = NULL,
  include_outside_option = FALSE
)

Arguments

data, id_col, alt_col, choice_col	As in wesml_weights.
n_per_alt	Either a single integer applied to every stratum, or a named integer vector of per-stratum counts (names matched to as.character(alt), covering all strata). Mutually exclusive with frac_per_alt.
frac_per_alt	Either a single fraction in [0, 1] applied to every stratum, or a named numeric vector of per-stratum fractions. Mutually exclusive with n_per_alt.
seed	Optional integer seed for reproducible sampling.
weight_name	Name of the attached weight column (default ".wesml_weight").
outside_opt_label, include_outside_option	As in wesml_weights (for an implicit outside good).

Details

Sampling is by choice situation (id), never by row: all alternative-rows of a sampled situation are kept together. Sampling is without replacement.

Value

A data.table subsample with the weight column appended and "Q", "H", and "choice_sampling" attributes (the last records the scheme, shares, quotas, and meat = "robust").

References

Manski, C. F. and Lerman, S. R. (1977). Econometrica 45(8), 1977-1988.

See Also

wesml_weights, run_mxlogit

Examples

library(data.table)
set.seed(1)
N <- 600L; J <- 3L
pop <- data.table(id = rep(seq_len(N), each = J), alt = rep(1:J, N))
pop[, x1 := rnorm(.N)]
pop[, w1 := rnorm(.N)]
pop[, choice := as.integer(seq_len(.N) == sample.int(.N, 1L)), by = id]

s <- sample_by_choice(pop, "id", "alt", "choice", n_per_alt = 50L, seed = 1L)
attr(s, "choice_sampling")$H   # realized sample shares
head(s[[".wesml_weight"]])

choicer documentation built on Sept. 5, 2026, 1:07 a.m.