Skip to contents

Collect predictions and LD scores from LDA results

Source: R/stat_lda.R
collect.stat_lda.Rd

Extract predictions, posterior probabilities, and optionally LD scores from an LDA result and add them to a tibble.

Usage

# S3 method for class 'stat_lda'
collect(x, data = NULL, retain = FALSE, fold = FALSE, ...)

Arguments

x

A stat_lda object

data

A tibble. If NULL, uses the original data from the LDA.

retain

How many LDs to retain:

  • FALSE (default): No LD scores, only predictions

  • TRUE: All LDs

  • Integer (e.g., 2): First N LDs

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

fold

How to add LD scores (only if retain != FALSE):

  • FALSE (default): Add as separate columns (LD1, LD2, ...)

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

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

...

Additional arguments (reserved)

Value

A tibble with predictions added:

  • pred: Predicted class (factor)

  • prob: Posterior probability of predicted class (numeric)

  • LD1, LD2, ... : LD scores (if retain is not FALSE)

Details

The function always adds cross-validated predictions (pred) and the posterior probability of the predicted class (prob). These come from leave-one-out cross-validation performed during model fitting.

Optionally, discriminant scores can be added via the retain parameter:

  • retain = FALSE: No LD scores (default, fastest)

  • retain = TRUE: All LD scores

  • retain = 2: First 2 LDs

  • retain = 0.95: LDs explaining 95% of between-group variance

When fold = TRUE or a character name, LD scores are stored as a list-column with class c("lda", "coe"), making them usable in downstream analyses.

Examples

if (FALSE) { # \dontrun{
lda <- boteft %>% stat_lda(type)

# Add predictions only (default)
collect(lda)

# Add predictions + all LD scores
collect(lda, retain = TRUE)

# Add predictions + first 2 LDs
collect(lda, retain = 2)

# Add predictions + LDs explaining 95% variance
collect(lda, retain = 0.95)

# Fold LD scores into list-column
collect(lda, retain = 2, fold = TRUE)
collect(lda, retain = 2, fold = "lda_scores")
} # }