The TwoGroupStats Class

Description

Compute row-by-row means and variances for a data matrix whose columns belong to two different groups of interest.

Usage

TwoGroupStats(data, classes, name = comparison, name1 = A, name2 = B)
## S4 method for signature 'TwoGroupStats':
as.data.frame(x, row.names=NULL, optional=FALSE)
## S4 method for signature 'TwoGroupStats':
summary(object, ...)
## S4 method for signature 'TwoGroupStats':
print(x, ...)
## S4 method for signature 'TwoGroupStats, missing':
plot(x, main=x@name, useLog=FALSE, ...)

Arguments

Details

This class was one of the earliest developments in our suite of tools to analyze microarrays. Its main purpose is to segregate out the preliminary computation of summary statistics on a row-by-row basis, along with a set of plots that could be generated automatically and used for quality control.

Creating Objects

As usual, objects can be created by new, but better methods are available in the form of the TwoGroupStats function. The inputs to this function are the same as those used for row-by-row statistical tests throughout the ClassComparison package; a detailed description can be found in the MultiTtest class.

One should note that this class serves as the front end to the SmoothTtest class, providing it with an interface that accepts ExpressionSet objects compatible with the other statistical tests in the ClassComparison package.

Slots

mean1:

The numeric vector of means in the first group

mean2:

The numeric vector of means in the second group

overallMean:

The numeric vector of overall row means

var1:

The numeric vector of variances in the first group

var2:

The numeric vector of variances in teh second group

overallVar:

The numeric vector of variances assuming the two groups have the same mean

pooledVar:

The numeric vector of row-by-row pooled variances, assuming the two groups have the same variance but different means

n1:

The number of items in the first group

n2:

The number of items in the first group

name1:

The name of the first group

name2:

The name of the second group

name:

The name of the object

Methods

as.data.frame(x, row.names=NULL, optional=FALSE)

Collect the numeric vectors from the object into a single dat fame, suitable for printing or exporting.

summary(object, ...)

Write out a summary of the object.

print(x, ...)

Print the object. (Actually, it only prints a summary, since the whole object is almost always more than you really want to see. If you insist on printing everything, use as.data.frame.)

plot(x, main=x@name, useLog=FALSE, ...)

This function actually produces six different plots of the data, so it is usually wrapped by a graphical layout command like par(mfrow=c(2,3)). The first two plots show the relation between the mean and standard deviation for the two groups separately; the third plot does the same for the overall mean and variance. The fourth plot is a Bland-Altman plot of the difference between the means against the overall mean. (In the microarray world, this is usually called an M-vs-A plot.) A loess fit is overlaid on the scatter plot, and points outside confidence bounds based on the fit are printed in a different color to flag them as highly variable. The fifth plot shows a loess fit (with confidence bounds) of the difference as a function of the row index (which often is related to the geometric position of spots on a microarray). Thus, this plot gives a possible indication of regions of an array where unusual things happen. The final plot compares the overall variances to the pooled variances.

Author(s)

Kevin R. Coombes <kcoombes@mdanderson.org>

References

Altman DG, Bland JM. Measurement in Medicine: the Analysis of Method Comparison Studies. The Statistician, 1983; 32: 307-317.

See Also

SmoothTtest, MultiTtest

Examples

bogus <- matrix(rnorm(30*1000, 8, 3), ncol=30, nrow=1000)
splitter <- rep(FALSE, 30)
splitter[16:30] <- TRUE

x <- TwoGroupStats(bogus, splitter)
summary(x)

opar<-par(mfrow=c(2,3), pch='.')
plot(x)
par(opar)

# cleanup
rm(bogus, splitter, x, opar)