Package for Generalized F-Statistics

Description

Implementation of several generalized F-statistics. The current version includes a generalized F-statistic based on the flexible isotonic/monotonic regression or order restricted hypothesis testing. Based on: Y. Lai (2011) <doi:10.1371/journal.pone.0019754>.

Details

Author(s)

Yinglei Lai

Maintainer: ylai@gwu.edu

A Flexible Order Restricted Hypothesis Testing

Description

These functions test the hypothesis regarding population means from ordered sample groups. Restrictions like a weakly/general/strongly isotonic/monotonic order as well as a lower bound for the location can be imposed on the population means. A partition of sample groups and the corresponding estimates of population means are also provided.

Usage

flexisoreg(y, x, lambda = 0, alpha.location = 1, alpha.adjacency = 0.5)
flexisoreg.stat(y, x, lambda = 0, alpha.location = 1, alpha.adjacency = 0.5)
flexmonoreg(y, x, lambda = 0, alpha.location = 1, alpha.adjacency = 0.5)
flexmonoreg.stat(y, x, lambda = 0, alpha.location = 1, alpha.adjacency = 0.5)

Arguments

Details

flexisoreg is used for flexible nondecreasing order restricted hypothesis testing. flexmonoreg is used for flexible nondecreasing or nonincreasing order restricted hypothesis testing. flexisoreg.stat and flexmonoreg.stat only return an F-statistic, which is convenient for multiple comparison.

Value

Note

Since the p-value of test has to be evaluated by permutation method, these functions will not return any p-value. For the permutation p-value of an individual test, see flexisoreg.pvalue and flexmonoreg.pvalue. For the pooled permutation p-values of multiple tests, see flexisoreg.poolpvalues and flexmonoreg.poolpvalues.

Author(s)

Yinglei Lai ylai@gwu.edu

References

Yinglei Lai (2007) A flexible order restricted hypothesis testing and its application to gene expression data. Technical Report

Examples

#generate ordinal group lables x
x <- runif(100)*6
x <- round(x,0)/3
#generate true values z
z <- round(x^2,0)
#generate observed values y
y <- z + rnorm(100)


#print default results
print(rbind(x,z,y))
print(flexisoreg(y,x))
print(flexisoreg.stat(y,x))
print(flexisoreg(y,0-x))
print(flexisoreg.stat(y,0-x))
print(flexmonoreg(y,x))
print(flexmonoreg.stat(y,x))


     #plots for illustration
     par(mfrow=c(2,3), mai=c(0.6, 0.6, 0.3, 0.1))
     plot(x,y, main="True Model",cex.axis=1.5, cex.lab=1.5, cex.main=1.5, cex=1.5)
     lines(x, z, type="p", pch=15, col="black", cex=2.5)

     results <- flexisoreg(y, x, lambda=1, alpha.location=0.05, alpha.adjacency=1)
     plot(x,y, main="Location Restriction",cex.axis=1.5, cex.lab=1.5, cex.main=1.5, cex=1.5)
     lines(x, results$estimate, type="p", pch=15, col="black", cex=2.5)

     results <- flexisoreg(y, x, lambda=1, alpha.location=0.05, alpha.adjacency=0.05)
     plot(x,y, main="Location and Strong Order Restrictions", 
     cex.axis=1.5, cex.lab=1.5, cex.main=1.5, cex=1.5)
     lines(x, results$estimate, type="p", pch=15, col="black", cex=2.5)

     results <- flexisoreg(y, x, lambda=0, alpha.location=1, alpha.adjacency=0.95)
     plot(x,y, main="Weak Order Restriction",cex.axis=1.5, cex.lab=1.5, cex.main=1.5, cex=1.5)
     lines(x, results$estimate, type="p", pch=15, col="black", cex=2.5)

     results <- flexisoreg(y, x, lambda=0, alpha.location=1, alpha.adjacency=0.5)
     plot(x,y, main="General Order Restriction",cex.axis=1.5, cex.lab=1.5, cex.main=1.5, cex=1.5)
     lines(x, results$estimate, type="p", pch=15, col="black", cex=2.5)

     results <- flexisoreg(y, x, lambda=0, alpha.location=1, alpha.adjacency=0.05)
     plot(x,y, main="Strong Order Restriction",cex.axis=1.5, cex.lab=1.5, cex.main=1.5, cex=1.5)
     lines(x, results$estimate, type="p", pch=15, col="black", cex=2.5)

Significance Assessment for the Flexible Order Restricted Hypothesis Testing

Description

These functions evaluate the p-values from an individual or multiple flexible order restricted hypothesis testing.

Usage

flexisoreg.pvalue(y, x, lambda=0, alpha.location=1, alpha.adjacency=0.5, B=100)
flexisoreg.poolpvalues(m, x, lambda=0, alpha.location=1, alpha.adjacency=0.5, B=100)
flexmonoreg.pvalue(y, x, lambda=0, alpha.location=1, alpha.adjacency=0.5, B=100)
flexmonoreg.poolpvalues(m, x, lambda=0, alpha.location=1, alpha.adjacency=0.5, B=100)

Arguments

Details

flexisoreg.pvalue and flexmonoreg.pvalue provide the permutation p-value for an individual flexible order restricted hypothesis testing. flexisoreg.poolpvalues and flexmonoreg.poolpvalues provide the pooled permutation p-values for multiple flexible order restricted hypothesis testing.

Value

flexisoreg.pvalue and flexmonoreg.pvalue return a permutation p-value. flexisoreg.poolpvalues and flexmonoreg.poolpvalues return a vector of pooled permutation p-values.

Note

These functions are used in conjunction with flexisoreg, flexisoreg.stat, flexmonoreg and flexmonoreg.stat.

Author(s)

Yinglei Lai ylai@gwu.edu

References

Yinglei Lai (2007) A flexible order restricted hypothesis testing and its application to gene expression data. Technical Report

Examples

#generate ordinal group lables x
x <- runif(100)*6
x <- round(x,0)/3
#generate true values z
z <- round(x^2,0)
#generate 6 vectors in a matrix for observed values, some noises and some not
m <- array(double(6*100), dim=c(6,100))
for(k in 1:3)
m[k,] <- rnorm(100)
for(k in 4:6)
m[k,] <- z + rnorm(100)


#print default results
par(mfrow=c(2,3))
for(k in 1:6){
print(paste("The ", k, "-th vector", sep=""))
y <- m[k,]
plot(x,y,main=k)
print(flexisoreg.stat(y,x))
print(flexisoreg.pvalue(y,x,B=20))
print(flexisoreg.stat(y,0-x))
print(flexisoreg.pvalue(y,0-x,B=20))
print(flexmonoreg.stat(y,x))
print(flexmonoreg.pvalue(y,x,B=20))
}

flexisoreg.poolpvalues(m, x, B=20)
flexmonoreg.poolpvalues(m, x, B=20)

Internal GeneF Functions

Description

Internal functions to support generalized F-statistics.

Usage

get.numbers(x)
t1p1(v, n)
t1p2(v, n1, n2)

Arguments

Value

Author(s)

Yinglei Lai ylai@gwu.edu