plot.cond {cond}

Generate Plots for an Approximate Conditional Inference Object

Description

Creates a set of plots for an object of class cond.

Usage

## S3 method for class 'cond'
plot(x = stop("nothing to plot"), from = x.axis[1], to = x.axis[n], 
     which = NULL, alpha = 0.05, add.leg = TRUE, loc.leg = FALSE,
     add.labs = TRUE, cex = 0.7, cex.lab = 1, cex.axis = 1, 
     cex.main = 1, lwd1 = 1, lwd2 = 2, lty1 = "solid", 
     lty2 = "dashed", col1 = "black", col2 = "blue", tck = 0.02, 
     las = 1, adj = 0.5, lab = c(15, 15, 5), ...)

Arguments

x

a cond object. This is assumed to be the result returned by the cond.glm function.

from

starting value for the x-axis range. The default value has been set by cond.glm.

to

ending value for the x-axis range. The default value has been set by cond.glm.

which

which plot should be printed. Admissible values are 2 to 8 corresponding to the choices in the menu below.

alpha

the level used to read off confidence intervals; default is 5%.

add.leg

if TRUE, a legend is added to each plot; default is TRUE.

loc.leg

if TRUE, position of the legend can be located by hand; default is FALSE.

add.labs

if TRUE, labels are added; default is TRUE.

cex, cex.lab, cex.axis, cex.main

character expansions relative to the standard size of the device to be used for printing text, labels, axes and main title. See par for details.

lwd1, lwd2

line width used to compare different curves in the same plot; default is lwd2 = 2 for higher order solutions and lwd1 = 1 for first order solutions.

lty1, lty2

line type used to compare different curves in the same plot; default is lty2 = "dashed" for the Wald statistic and lty1 = "solid" for the remaining first- and higher order statistics.

col1, col2

colors used to compare different curves in the same plot; default is col2 = "blue" for higher order solutions, and col1 = "black" for the remaining first order statistics.

tck, las, adj, lab

further graphical parameters. See par for details.

...

optional graphical parameters; see par for details.

Details

Several plots are produced for an object of class cond. A menu lists all the plots that can be produced. They may be one or all of the following ones:

 Make a plot selection (or 0 to exit)

1:plot: All
2:plot: Profile and modified profile log likelihoods
3:plot: Profile and modified profile likelihood ratios
4:plot: Profile and modified likelihood roots
5:plot: Modified and continuity corrected likelihood roots
6:plot: Lugannani-Rice approximations
7:plot: Confidence intervals
8:plot: Diagnostics based on INF/NP decomposition

Selection:
  

If no nuisance parameters are presented, a subset of the above pictures is produced. More details on the implementation are given in Brazzale (1999, 2000).

This function is a method for the generic function plot() for class cond. It can be invoked by calling plot or directly plot.cond for an object of the appropriate class.

Value

A plot is created on the current graphics device.

Side Effects

The current device is cleared. When add.leg = TRUE, a legend is added to each plot, and if loc.leg = TRUE, it can be set by the user. All screens are closed, but not cleared, on termination of the function.

Note

The diagnostic plots only represent a preliminary version and need further development.

The two panels on the right trace the information and nuisance correction terms, INF and NP, against the likelihood root statistic. These are generally smooth functions and used to approximate the information and nuisance parameter aspects as a function of the parameter of interest, as shown in the two panels on the left. This procedure has the advantage of largely eliminating the numerical instabilities that affect the statistics around the MLE. The circles in the two leftmost panels represent the limit of INF and NP at the MLE calculated exactly using numerical derivatives. All four pictures are intended to give an idea of the order of magnitude of the two correction terms while trying to deal with the numerical problems that likely occur for these kinds of data.

More details can be found in Brazzale (2000, Appendix B.2).

References

Brazzale, A. R. (1999) Approximate conditional inference for logistic and loglinear models. J. Comput. Graph. Statist., 8, 1999, 653–661.

Brazzale, A. R. (2000) Practical Small-Sample Parametric Inference, Ph.D. Thesis N. 2230, Department of Mathematics, Swiss Federal Institute of Technology Lausanne.

See Also

cond.glm, cond.object, summary.cond

Examples

## Crying Babies Data
data(babies)
babies.glm <- glm(formula = cbind(r1, r2) ~ day + lull - 1, 
                  family = binomial, data = babies)
babies.cond <- cond(object = babies.glm, offset = lullyes)
## Not run: 
plot(babies.cond)

## End(Not run)

## Urine Data 
data(urine)
urine.glm <- glm(r ~ I(gravity * 100) + ph + osmo + conduct + urea + calc, 
                 family = binomial, data = urine)
urine.cond <- cond(urine.glm, I(gravity * 100))
plot(urine.cond, which=4)