regr1.plot: plot x and y, with optional straight line fit and display of...

In HH: Statistical Analysis and Data Display: Heiberger and Holland

plot x and y, with optional straight line fit and display of squared residuals

Description

Plot x and y, with optional fitted line and display of squared residuals. By default the least squares line is calculated and used. Any other straight line can be specified by placing its coefficients in coef.model. Any other fitted model can be calculated by specifying the model argument. Any other function of one variable can be specified in the alt.function argument. At most one of the arguments model, coef.model, alt.function can be specified.

Usage

regr1.plot(x, y,
           model=lm(y~x),
           coef.model,
           alt.function,
           main="put a useful title here",
           xlab=deparse(substitute(x)),
           ylab=deparse(substitute(y)),
           jitter.x=FALSE,
           resid.plot=FALSE,
           points.yhat=TRUE,
           pch=16,
           ..., length.x.set=51,
           x.name,
           pch.yhat=16,
           cex.yhat=par()$cex*.7,
           err=-1)

Arguments

x	x variable
y	y variable
model	Defaults to the simple linear model lm(y ~ x). Any model object with one x variable, such as the quadratic lm(y ~ x + I(x^2)) can be used.
coef.model	Defaults to the coefficients of the model argument. Other intercept and slope coefficients for a straight line (for example, c(3,5)) can be entered to illustrate the sense in which they are not "least squares".
alt.function	Any function of a single argument can be placed here. For example, alt.function=function(x) {3 + 2*x + 3*x^2}. All coefficients must be specified.
main, xlab, ylab	arguments to plot.
jitter.x	logical. If TRUE, the x is jittered before plotting. Jittering is often helpful when there are multiple y-values at the same level of x.
resid.plot	If FALSE, then do not plot the residuals. If "square", then call resid.squares to plot the squared residuals. If TRUE (or anything else), then call resid.squares to plot straight lines for the residuals.
points.yhat	logical. If TRUE, the predicted values are plotted.
...	other arguments.
length.x.set	number of points used to plot the predicted values.
x.name	If the model argument used a different name for the independent variable, you might need to specify it.
pch	Plotting character for the observed points.
pch.yhat	Plotting character for the fitted points.
cex.yhat	cex for the fitted points.
err	The default -1 suppresses warnings about out of bound points.

Note

This plot is designed as a pedagogical example for introductory courses. When resid.plot=="square", then we actually see the set of squares for which the sum of their areas is minimized by the method of "least squares".

Author(s)

Richard M. Heiberger <[e-mail]>

References

Heiberger, Richard M. and Holland, Burt (2015). Statistical Analysis and Data Display: An Intermediate Course with Examples in R. Second Edition. Springer-Verlag, New York. https://link.springer.com/book/10.1007/978-1-4939-2122-5

Smith, W. and Gonick, L. (1993). The Cartoon Guide to Statistics. HarperCollins.

See Also

resid.squares

Examples

data(hardness)

## linear and quadratic regressions
hardness.lin.lm  <- lm(hardness ~ density,                data=hardness)
hardness.quad.lm <- lm(hardness ~ density + I(density^2), data=hardness)

anova(hardness.quad.lm)  ## quadratic term has very low p-value

par(mfrow=c(1,2))

regr1.plot(hardness$density, hardness$hardness,
           resid.plot="square",
           main="squared residuals for linear fit",
           xlab="density", ylab="hardness",
           points.yhat=FALSE,
           xlim=c(20,95), ylim=c(0,3400))

regr1.plot(hardness$density, hardness$hardness,
           model=hardness.quad.lm,
           resid.plot="square",
           main="squared residuals for quadratic fit",
           xlab="density", ylab="hardness",
           points.yhat=FALSE,
           xlim=c(20,95), ylim=c(0,3400))

par(mfrow=c(1,1))

HH documentation built on May 29, 2024, 6:24 a.m.