Traceplot method for hierarchical Bayes fits

Overlaid per-chain traceplots (base graphics, no new dependency) for a choicer_hmnl / choicer_hmnp fit's population coefficients (b), delta mean-function coefficients (theta), alternative-effect variance (sigma_d2), and, opt-in, a representative subset of the (potentially ~200-column) alternative effects (delta).

Usage

# S3 method for class 'choicer_hb'
traceplot(object, block = c("b", "theta", "sigma_d2"), which = NULL, ...)

Arguments

object

A choicer_hmnl or choicer_hmnp fit.

block

Character vector, any non-empty subset of c("b", "theta", "sigma_d2", "delta") (HMNP fits additionally accept "sigma2", the raw non-identified shock-variance chain). Default plots b, theta, sigma_d2 (not delta, which can have ~200 columns).

which

Only consulted when "delta" is in block. NULL (default) auto-selects a representative subset (the 3 highest rank-R-hat plus 3 lowest ESS_bulk alternatives, deduplicated, capped at 6). If supplied, a character vector of alternative labels or an integer vector of column indices.

...

Additional arguments (ignored).

Value

object, invisibly.

Examples

# \donttest{
sim <- simulate_hmnl_data(N = 60, T = 2, J = 4, seed = 1)
fit <- suppressWarnings(run_hmnlogit(sim$data, "task", "alt", "choice", c("x1", "x2"),
                    person_col = "pid",
                    mcmc = list(R = 300, burn = 100), chains = 2))
#> MCMC run time 0h:0m:0.02s
traceplot(fit)

# }