Transduce model positions to coefficients and shapes

Predict coefficients and reconstruct shapes at specified positions in model space.

Usage

# S3 method for class 'stat_lda'
transduce(object, positions)

# S3 method for class 'stat_pca'
transduce(object, positions)

transduce(object, positions)

Arguments

object

A stat_* object (stat_pca, stat_lda, stat_lm, ...)

positions

A tibble with position coordinates. Column names must match the model's coordinate system (e.g., PC1, PC2 for PCA). Each row represents one position where coefficients will be reconstructed.

Value

A tibble with columns:

• Position column(s): from input positions tibble

• Coefficient column(s): coe (or coe_A, coe_B, ...) with proper classes

• Inverse shape column(s): coe_i (or coe_A_i, coe_B_i, ...) as matrices

Details

transduce() reconstructs coefficient vectors at specified positions in model space and applies inverse transformations to obtain shapes.

The positions tibble defines where to reconstruct shapes. For PCA, columns should be named PC1, PC2, etc. Unspecified PCs default to 0 (mean position).

Shapes are reconstructed using the appropriate inverse method based on coefficient class (eft_i, rft_i, opoly_i, etc.).

See also

tidyr::expand_grid() for creating position grids

Examples

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

# Single axis
pos <- tibble(PC1 = c(-2, 0, 2))
trans <- transduce(pca, pos)

# Two axes (manual grid)
pos <- tibble(
  PC1 = c(-2, 0, 2, -2, 0, 2),
  PC2 = c(-1, -1, -1, 1, 1, 1)
)
trans <- transduce(pca, pos)

# Or use expand_grid for convenience
pos <- expand_grid(PC1 = c(-2, 0, 2), PC2 = c(-1, 1))
trans <- transduce(pca, pos)

# Access reconstructed shapes
plot(trans$coe_i[[1]], type = "l")
} # }