Skip to contents

Predict method for PCA

Source: R/stat_pca.R
predict.stat_pca.Rd

Project new data onto principal components from a fitted PCA model.

Usage

# S3 method for class 'stat_pca'
predict(object, newdata, retain = NULL, fold = FALSE, .collect = TRUE, ...)

Arguments

object

A stat_pca object

newdata

A tibble with the same predictor columns as training data

retain

How many PCs to return:

  • NULL (default): All PCs

  • Integer (e.g., 5): First N PCs

  • Numeric 0-1 (e.g., 0.95): PCs explaining this proportion of variance

fold

How to return PC scores:

  • FALSE (default): Add as separate columns (PC1, PC2, ...)

  • TRUE: Fold into single list-column named "pca"

  • Character: Fold into single list-column with this name

.collect

Logical. Should predictions be added to newdata (TRUE, default) or returned as a standalone tibble (FALSE)?

...

Additional arguments (reserved)

Value

If .collect = TRUE, returns newdata with PC scores added. If .collect = FALSE, returns a tibble with identifier columns (non-coe) and PC scores only.

Examples

if (FALSE) { # \dontrun{
# Train PCA
pca <- boteft %>% stat_pca()

# Predict on new data
new_scores <- predict(pca, new_data)

# Pipe the model
new_scores <- training %>%
  stat_pca() %>%
  predict(testing)
} # }