Skip to contents

Model-Fit Scalars in One Call

Source: R/likelihood.R
model_fit.Rd

Bundles the standard goodness-of-fit scalars for a fitted tree into one tidy row: logLik, the parameter count df, the observation count nobs, AIC, BIC, and perplexity. A one-call replacement for logLik(); nobs(); AIC(); BIC(); perplexity().

Usage

model_fit(tree, newdata = NULL)

Arguments

tree

A transitiontrees or transitiontrees_group.

newdata

Optional sequence data. If supplied, the scalars are evaluated on it (held-out); if NULL (default), in-sample.

Value

A one-row data.frame (one row per group for a transitiontrees_group) with columns logLik, df, nobs, AIC, BIC, perplexity.

Details

With newdata, every scalar is computed out-of-sample (AIC/BIC use the held-out deviance with the model's training df). A transitiontrees_group returns one row per group, tagged with a leading group column.

See also

perplexity, logLik.transitiontrees.

Examples

# \donttest{
seqs <- replicate(60, sample(c("A","B","C"), 12, replace = TRUE),
                  simplify = FALSE)
tree <- context_tree(seqs, max_depth = 2L, min_count = 3L)
model_fit(tree)
#>    logLik df nobs     AIC     BIC perplexity
#> 1 -781.96 26  720 1615.92 1734.98   2.962565
# }