trans.plot {pda}R Documentation

Diagnostic plot to find power transform to remove interaction. See Emerson chapter in Hoaglin, Mosteller and Tukey (1991, ch. 13E) Fundamentals of Exploratory Analysis of Variance (Wiley);or see Mosteller and Tukey (1977, ch. 9E,F) Exploratory Data Analysis (Addison-Wesley).

Description

Usage

trans.plot(fit, predictors=all.predictors(fit)[1:2],
  terms=predict(fit, type = "terms"), xlab, ylab, ...)

Arguments

fit linear model (lm) fit
predictors predictors to compare. They must be in the model, as well as theirinteraction in the specified order.
terms estimates of model terms
xlab,ylab x and y axis labels
... Plot parameters

Value

Power = 1 - slope estimate. SE comes from simple linear regression and is at best advisory. Consider tranformation $y^p$ = exp(p*log(y)), or transform log(y) if p=0. In practice, examine nearest half integer.

See Also

effect.plot.

Examples

## Not run: 
# Bekk data from Hoaglin, Mosteller and Tukey (1991)
bekk.fit <- lm(smooth~lab*mat, bekk)
effect.plot(bekk.fit)trans.plot(bekk.fit)
bekk.lfit <- lm(log(smooth)~lab*mat, bekk)
effect.plot(bekk.lfit)
trans.plot(bekk.lfit)
## End(Not run)

[Package pda version 1.2-7 Index]