predict.drc: Prediction

In DoseResponse/drc: Analysis of Dose-Response Data

Description

Predicted values for models of class 'drc'.

Usage

  ## S3 method for class 'drc'
predict(object, newdata, se.fit = FALSE, 
  interval = c("none", "confidence", "prediction", "ssd"), 
  level = 0.95, na.action = na.pass, od = FALSE, vcov. = vcov, 
  ssdSEfct = NULL, constrain = TRUE, checkND = TRUE, ...)

Arguments

object	an object of class 'drc'.
newdata	An optional data frame in which to look for variables with which to predict. If omitted, the fitted values are used.
se.fit	logical. If TRUE standard errors are required.
interval	character string. Type of interval calculation: "none", "confidence", "prediction", or "ssd".
level	Tolerance/confidence level.
na.action	function determining what should be done with missing values in 'newdata'. The default is to predict 'NA'.
od	logical. If TRUE adjustment for over-dispersion is used.
vcov.	function providing the variance-covariance matrix. vcov is the default, but sandwich is also an option (for obtaining robust standard errors).
ssdSEfct	specifies the function for interpolating standard errors between observed standard errors. The default is linear interpolation on log-log scale (back-transformed). See Details for more explanation.
constrain	logical. If TRUE (default) predicted values are truncated within meaningful limits, i.e., 0 and, possibly, 1.

checkND	logical indicating whether or not names in "newdata" data frame match the names in the original data frame (used for fitting the model). Default is TRUE.
...	further arguments passed to or from other methods.

Details

For the built-in log-logistic, log-normal, and Weibull-type models standard errors and confidence/prediction intervals can be calculated. For other built-in models it may not yet be implemented (drop us an e-mail if you need them).

The function for interpolating standard errors of estimates, which may be used when fitting an SSD, should have 3 arguments: observed estimates and corresponding standard errors and future estimates and should return interpolated standard errors corresponding to the future estimates provided.

Value

A matrix with as many rows as there are dose values provided in 'newdata' or in the original dataset (in case 'newdata' is not specified) and, at most, 4 columns containing fitted, standard errors, lower and upper limits of confidence/prediction intervals.

Author(s)

Christian Ritz

See Also

For details are found in the help page for predict.lm.

Examples

## Fitting a model
spinach.model1 <- drm(SLOPE~DOSE, CURVE, data = spinach, fct = LL.4())

## Predicting values a dose=2 (with standard errors)
predict(spinach.model1, data.frame(dose=2, CURVE=c("1", "2", "3")), se.fit = TRUE)

## Getting confidence intervals
predict(spinach.model1, data.frame(dose=2, CURVE=c("1", "2", "3")), 
interval = "confidence")

## Getting prediction intervals
predict(spinach.model1, data.frame(dose=2, CURVE=c("1", "2", "3")), 
interval = "prediction")

DoseResponse/drc documentation built on May 7, 2021, 4:55 p.m.