ancova: Robust ANCOVA

Description

This function computes robust ANCOVA for 2 independent groups and one covariate. It compares trimmed means. No parametric assumption (e.g. homogeneity) is made about the form of the regression lines. A running interval smoother is used. A bootstrap version which computes confidence intervals using a percentile t-bootstrap is provided as well.

Usage

ancova(formula, data, tr = 0.2, fr1 = 1, fr2 = 1, pts = NA, ...)
ancboot(formula, data, tr = 0.2, nboot = 599, fr1 = 1, fr2 = 1, pts = NA, ...)

Value

Returns an object of class ancova containing:

evalpts: covariate values (including points close to these values) where the test statistic is evaluated
n1: number of subjects at evaluation point (first group)
n2: number of subjects at evaluation point (first group)
trDiff: trimmed mean differences
se: standard errors for trimmed mean differences
ci.low: lower confidence limit for trimmed mean differences
ci.hi: upper confidence limit for trimmed mean differences
test: values of the test statistic
crit.vals: critical values
p.vals: p-values
fitted.values: fitted values from interval smoothing
call: function call

Arguments

formula: an object of class formula.
data: an optional data frame for the input data.
tr: trim level for the mean.
fr1: values of the span for the first group (1 means unspecified)
fr2: values of the span for the second group (1 means unspecified)
pts: can be used to specify the design points where the regression lines are to be compared; if NA design points are chosen.
nboot: number of bootstrap samples
...: currently ignored.

References

Wilcox, R. (2012). Introduction to Robust Estimation and Hypothesis Testing (3rd ed.). Elsevier.

Examples

Run this code

head(invisibility)
ancova(mischief2 ~ cloak + mischief1, data = invisibility)

## specifying covariate evaluation points
ancova(mischief2 ~ cloak + mischief1, data = invisibility, pts = c(3, 4, 8, 1))

## bootstrap version
ancboot(mischief2 ~ cloak + mischief1, data = invisibility)