predict.vem_fit: Predict Method for VEM Fits

In fda.vi: Functional Data Analysis using Variational Inference

Predict Method for VEM Fits

Description

Returns posterior mean curve estimates from a vem_fit object. Active basis functions are selected by applying a 0.5 probability threshold on the posterior inclusion probabilities. If newdata is supplied, a new basis matrix is constructed at those time points; otherwise the original fitted time points are used. Predictions are automatically back-transformed if the model was fitted with center = TRUE or scale = TRUE.

Usage

## S3 method for class 'vem_fit'
predict(object, newdata = NULL, ...)

Arguments

object	A vem_fit object from vem_fit.
newdata	Optional numeric vector of new time points at which to evaluate the fitted curves. Must lie within the original domain range(Xt). If NULL, predictions are returned at the original Xt.
...	Currently unused.

Value

A list of length m. Each element is a numeric vector of predicted values on the original (back-transformed) scale.

References

da Cruz, A. C., de Souza, C. P. E., & Sousa, P. H. T. O. (2024). Fast Bayesian basis selection for functional data representation with correlated errors. arXiv:2405.20758. https://arxiv.org/abs/2405.20758

See Also

vem_fit, plot.vem_fit, coef.vem_fit

Examples

  data(toy_curves)
  fit <- vem_fit(y = toy_curves$y, Xt = toy_curves$Xt, K = 8)

  # Predictions at original time points
  preds <- predict(fit)
  length(preds)       # 3 — one vector per curve

  # Predictions at a denser grid
  Xt_new <- seq(0, 1, length.out = 200)
  preds_dense <- predict(fit, newdata = Xt_new)

  # Plot observed vs predicted for curve 1
  plot(toy_curves$Xt, toy_curves$y[[1]],
       pch = 16, cex = 0.6, col = "grey50",
       xlab = "t", ylab = "y")
  lines(Xt_new, preds_dense[[1]], col = "firebrick", lwd = 2)

fda.vi documentation built on June 20, 2026, 5:06 p.m.