Evaluate the log-likelihood of behavioral data

Compute the log-likelihood of prepared behavioral data under one or more candidate parameter sets.

Usage

log_likelihood(
  context,
  data,
  parameters,
  ok = NULL,
  expand = NULL,
  min_ll = log(1e-10),
  ...
)

# S3 method for class 'accumulatr_context'
log_likelihood(
  context,
  data,
  parameters,
  ok = NULL,
  expand = NULL,
  min_ll = log(1e-10),
  ...
)

# Default S3 method
log_likelihood(context, ...)

Arguments

context

Context created with `make_context()`.

data

Prepared data created with `prepare_data()`.

parameters

A parameter data frame, or a list of parameter data frames.

ok

Logical vector marking which trials should contribute to the likelihood. Trials marked `FALSE` are assigned `min_ll`.

expand

Optional index vector used to expand compressed trial-level results back to the original trial count.

min_ll

Minimum log-likelihood value used for excluded or impossible trials.

...

Unused; for S3 compatibility.

Value

A numeric vector of log-likelihood values.

Examples

spec <- race_spec()
spec <- add_accumulator(spec, "A", "lognormal")
spec <- add_outcome(spec, "A_win", "A")
structure <- finalize_model(spec)
params_df <- build_param_matrix(
  spec,
  c(A.m = 0, A.s = 0.1, A.q = 0, A.t0 = 0),
  n_trials = 2
)
data_df <- simulate(structure, params_df, seed = 1)
prepared <- prepare_data(structure, data_df)
ctx <- make_context(structure)
log_likelihood(ctx, prepared, list(params_df))
#> [1] 2.481128