SAGx (version 1.46.0)

JT.test: Jonckheere-Terpstra trend test

Description

The test is testing for a monotone trend in terms of the class parameter. The number of times that an individual of a higher class has a higher gene expression forms a basis for the inference.

Usage

JT.test(data, class, labs = NULL, alternative = c("two-sided", "decreasing", "increasing"), ties = FALSE)

Arguments

data

A matrix with genes in rows and subjects in columns

class

the column labels, if not an ordered fctor it will be redefined to be one.

labs

the labels of the categories coded by class

alternative

two-sided, decreasing or increasing

ties

Adjustment for ties

Value

• an object of class JT-test, which extends the class htest, and includes the following slots
• statistic
  the observed JT statistic
• parameter
  the null hypothesis parameter, if other value than 0.
• p.value
  the p-value for the two-sided test of no trend.
• method
  Jonckheere-Terpstra
• alternative
  The relations between the levels: decreasing, increasing or two-sided
• data.name
  the name of the input data
• median1 ... mediann
  the medians for the n groups
• trend
  the rank correlation with category
• S1
  Predictive strength

Details

Assumes that groups are given in increasing order, if the class variable is not an ordered factor, it will be redefined to be one. The p-value is calculated through a normal approximation. The implementation owes to suggestions posted to R list. The definition of predictive strength appears in Flandre and O'Quigley.

References

Lehmann, EH (1975) Nonparametrics: Statistical Methods Based on Ranks p. 233. Holden Day Flandre, Philippe and O'Quigley, John, Predictive strength of Jonckheere's test for trend: an application to genotypic scores in HIV infection, Statistics in Medicine, 2007, 26, 24, 4441-4454

Examples

# Enter the data as a vector
A <- as.matrix(c(99,114,116,127,146,111, 125,143,148,157,133,139, 149, 160, 184))
# create the class labels
g <- c(rep(1,5),rep(2,5),rep(3,5))
# The groups have the medians
tapply(A, g, median)
# JT.test indicates that this trend is significant at the 5% level
JT.test(data = A, class = g, labs = c("GRP 1", "GRP 2", "GRP 3"), alternative = "two-sided")