Predictions from a fitted model — predict.femtofit • femtograd

Generate predictions from a femtofit object using the prediction function supplied at fit time.

Usage

# S3 method for class 'femtofit'
predict(object, newdata, ...)

Arguments

object: A femtofit object.
newdata: New data for prediction. The format depends on the prediction function provided to fit().
...: Additional arguments passed to the prediction function.

Value

Predictions, as returned by the prediction function.

Details

For predict() to work, a predict_fn must have been provided when calling fit(). The prediction function should have signature function(params, newdata, ...) where params is a named list of parameter values.

Examples

if (FALSE) { # \dontrun{
# Linear regression example
x <- 1:10
y <- 2 + 3*x + rnorm(10, sd = 0.5)

# Fit with prediction function
result <- fit(
  function(a, b, log_sigma) {
    sigma <- exp(log_sigma)
    predicted <- a + b * x
    loglik_normal(predicted, sigma, y)
  },
  params = c(a = 0, b = 1, log_sigma = 0),
  predict_fn = function(params, newdata) {
    params$a + params$b * newdata
  }
)

# Predict for new x values
predict(result, newdata = c(11, 12, 13))
} # }