variable_importance {cytominer}

Measure variable importance.

Description

variable_importance measures importance of variables based on specified methods.

Usage

variable_importance(
  sample,
  variables,
  operation = "replicate_correlation",
  ...
)

Arguments

sample

tbl containing sample used to estimate parameters.

variables

character vector specifying observation variables.

operation

optional character string specifying method for computing variable importance. Currently, only "replicate_correlation" (default) is implemented.

...

arguments passed to variable importance operation.

Value

data frame containing variable importance measures.

Examples

set.seed(123)
x1 <- rnorm(10)
x2 <- x1 + rnorm(10) / 100
y1 <- rnorm(10)
y2 <- y1 + rnorm(10) / 10
z1 <- rnorm(10)
z2 <- z1 + rnorm(10) / 1

batch <- rep(rep(1:2, each = 5), 2)

treatment <- rep(1:10, 2)

replicate_id <- rep(1:2, each = 10)

sample <-
  tibble::tibble(
    x = c(x1, x2), y = c(y1, y2), z = c(z1, z2),
    Metadata_treatment = treatment,
    Metadata_replicate_id = replicate_id,
    Metadata_batch = batch
  )

head(sample)

# `replicate_correlation`` returns the median, min, and max
# replicate correlation (across batches) per variable
variable_importance(
  sample = sample,
  variables = c("x", "y", "z"),
  operation = "replicate_correlation",
  strata = c("Metadata_treatment"),
  replicates = 2,
  split_by = "Metadata_batch",
  replicate_by = "Metadata_replicate_id",
  cores = 1
)