Varying coefficients

The coefficient functions of a model fitted with vc() terms, evaluated at each observation. A forest has no coefficient vector, so a fit with no vc() term has no coefficient to report and this errors rather than returning anything.

Usage

# S3 method for class 'bartisan_fit'
coef(object, newdata = NULL, draws = FALSE, ...)

Arguments

object

a <bartisan_fit> object; the output of a call to bartisan(), fitted with at least one vc() term.

newdata

optional; a data frame at which to evaluate the coefficients. Default is NULL to use the data the model was fitted to.

draws

logical; whether to return every posterior draw of each coefficient rather than its posterior mean at each observation. Default is FALSE to return the posterior means.

...

not used.

Value

With draws = FALSE, a matrix with one row per observation and one column per coefficient. With draws = TRUE, a named list of draws-by-observations matrices, one per coefficient.

Details

The control function is not among them. It is the surface at the value each covariate was centered on, which is a prediction rather than a coefficient; predict(object) is what reports predictions.

For a factor the coefficients are recentered to sum to zero across its levels, which is what makes them the deviations they are reported as. The symmetric coding carries one spare function-valued dimension, so this is exact rather than an approximation, and it is the reason a factor's reference level is a choice made here rather than at fitting time.

See also

vc() for declaring a varying coefficient; variable_importance() for which predictors a forest uses at all

Examples

data("rhc")
set.seed(123)

# The effect of catheterization is allowed to vary with the other
# predictors, so its coefficient is a function rather than a number
fit <- bartisan(death ~ age + aps + surv2m + vc(rhc), data = rhc,
                num_trees = 10, num_burn = 50, num_draws = 50,
                verbose = FALSE)
#> ℹ Using `family = binomial()`.
#> ℹ Set `family` to choose another, which also silences this message.

# One coefficient per patient, on the link scale
head(coef(fit))
#>            rhc
#> [1,] 0.4263262
#> [2,] 0.2144616
#> [3,] 0.2268549
#> [4,] 0.2412446
#> [5,] 0.2357062
#> [6,] 0.1668163

# How much it varies across patients
quantile(coef(fit)[, "rhc"])
#>        0%       25%       50%       75%      100% 
#> 0.1211476 0.2293972 0.3558278 0.4252599 0.5598030