predict.VFP: Predict Method for Objects of Class 'VFP'.

In VFP: Variance Function Program

Predict Method for Objects of Class 'VFP'.

Description

Predictions are made for the variance (type="vc"), standard deviation ("sd") or coefficient of variation ("cv") and their corresponding confidence intervals. The latter are calculated primarily on the variance scale and then transformed to the other scales, if required.

Usage

## S3 method for class 'VFP'
predict(
  object,
  model.no = NULL,
  newdata = NULL,
  alpha = 0.05,
  dispersion = NULL,
  type = c("vc", "sd", "cv"),
  CI.method = c("chisq", "t", "normal"),
  use.log = FALSE,
  ...
)

Arguments

object	(object) of class "VFP"
model.no	(integer) specifying a fitted model stored in 'object'
newdata	(numeric) optionally, a vector specifying mean-values for which predictions on the user-defined scale ('type') are requested. If omitted, fitted values will be returned.
alpha	(numeric) value specifying the 100 x (1-alpha)% confidence interval of predicted values
dispersion	(numeric) NULL = the dispersion will be set =1 (should usually not be changed; For the Saddler model, the dispersion is 1.), numeric value = the dispersion parameter will be used as specified
type	(character) specifying on which scale the predicted values shall be returned, possible are "vc" = variance, "sd"=standard deviation, "cv"=coefficient of variation
CI.method	(character) one of "t", "normal", "chisq" specifying which CI-method to use
use.log	(logical) TRUE = X- and Y-axis will be log-transformed
...	additional parameters passed forward to function predict.gnm

Value

(data.frame) with numeric variables:

Mean	value at which predictions were requested
Fitted	prediction at 'Mean'
SE	standard error of prediction
Scale	residual scale
LCL	lower confidence limit of the 100x(1-'alpha')% CI
UCL	upper confidence limit of the 100x(1-'alpha')% CI

Examples

library(VCA)
data(VCAdata1)
lst <- anovaVCA(y~(device+lot)/day/run, VCAdata1, by="sample")
mat <- getMat.VCA(lst)		# automatically selects "total"
res <- fit.vfp(model.no=1:10, Data=mat)
predict(res)
predict(res, dispersion=0.95)

VFP documentation built on Nov. 10, 2022, 5:12 p.m.