Predict from a fitted model

Predict from a fitted model

Usage

# S3 method for class 'brs'
predict(
  object,
  newdata = NULL,
  type = c("response", "link", "precision", "variance", "quantile", "score",
    "expected_score"),
  at = 0.5,
  ...
)

Arguments

object

A fitted "brs" object.

newdata

Optional data frame for prediction.

type

Prediction type: "response" (default; the mean \(E[Y] = a / (a + b)\)), "link" (linear predictor of the first parameter), "precision" (second parameter on its own scale), "variance", "quantile", "score" or "expected_score". "score" is the latent continuous score of the fit's interval at \(y^* = E[Y]\): \(K y^*\) on \((0, K)\) for "mid", \((K + 1) y^*\) on \((0, K + 1)\) for "right" and \((K + 1) y^* - 1\) on \((-1, K)\) for "left" (it can be negative); under "right"/"left" it is about 0.5 above/below the expected recorded score, because a recorded score is the lower/upper end of its latent interval. "expected_score" is the expected recorded score \(\sum_s s\, P(S = s)\) (brs_predict_scoreprob), the same for "right" and "left", and a proper expectation only when the cells partition \([0, 1]\).

at

Numeric vector of probabilities for quantile predictions (default 0.5).

...

Currently ignored.

Value

Numeric vector or matrix.

See also

brs, fitted.brs, brs_predict_scoreprob

Examples

# \donttest{
dat <- data.frame(
  y = c(
    0, 5, 20, 50, 75, 90, 100, 30, 60, 45,
    10, 40, 55, 70, 85, 25, 35, 65, 80, 15
  ),
  x1 = rep(c(1, 2), 10)
)
prep <- brs_prep(dat, ncuts = 100)
#> brs_prep: n = 20 | exact = 0, left = 1, right = 1, interval = 18
fit <- brs(y ~ x1, data = prep)
head(predict(fit))
#> [1] 0.5087226 0.4538041 0.5087226 0.4538041 0.5087226 0.4538041
head(predict(fit, type = "precision"))
#> [1] 0.4030159 0.4030159 0.4030159 0.4030159 0.4030159 0.4030159
newdat <- data.frame(x1 = c(1, 2))
predict(fit, newdata = newdat)
#> [1] 0.5087226 0.4538041
# }