simulate_nl_data: Simulate nested logit data

In choicer: Discrete Choice Models for Economic Applications

Simulate nested logit data

Description

Generates synthetic choice data with nested logit probabilities computed analytically (log-sum-exp over inclusive values), then samples choices from the implied multinomial. The outside option (j = 0) sits in a singleton nest with lambda = 1.

Usage

simulate_nl_data(
  N = 10000,
  beta = c(1.5, -0.8),
  delta = c(`1` = 0.5, `2` = 0.3, `3` = -0.2, `4` = -0.5, `5` = 0.4),
  nests = list(c(1, 2), c(3, 4, 5)),
  lambdas = c(0.8, 0.2),
  seed = 123
)

Arguments

N	Number of choice situations.
beta	Fixed coefficients for covariates ⁠X, W⁠ (length 2 by default).
delta	Named numeric vector of ASCs for inside alternatives.
nests	List of integer vectors defining nest membership for inside alternatives.
lambdas	Numeric vector of dissimilarity parameters, one per nest.
seed	Random seed (NULL skips set.seed()).

Value

A choicer_sim object. true_params includes beta, delta, lambdas; settings includes the nest_structure. The returned data retains a nest column (integer, with 0L for the outside option) for convenient use with run_nestlogit().

Note

Unlike simulate_mnl_data() and simulate_mxl_data(), this function does not expose outside_option or vary_choice_set flags. The outside option (j = 0) is always present as a singleton nest with lambda = 1, and every individual faces the full set of inside alternatives. Add these flags if downstream use cases need them.

Examples

sim <- simulate_nl_data(N = 2000, seed = 123)
print(sim)

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