R/signt.R

In musto101/wilcox_R: What the package does (short line)

signt <-
function(x,y=NULL,alpha=.05,AC=TRUE,SD=FALSE){
#
#  Do a sign test on data in x and y
#  If y=NA, assume x is a matrix with
#  two columns or has list mode.
#
#  Returns n, the original sample size
#  N, number of paired observations that are not equal to one another.
#  phat, an estimate of p, the probability that x<y.
#  ci, a confidence interval for p
#
#  Pratt's method is used by default.
#  AC=T, the Agresti-Coull method is used instead.
#  SD=TRUE, the Schilling-Doi method is used, which returns a confidence
#           interval but no p-value.
#
if(is.null(y[1])){
if(ncol(as.matrix(x))!=2)stop('y is null so x should be a matrix or data frame with two columns')
if(is.matrix(x)||is.data.frame(x))dif<-x[,1]-x[,2]
if(is.list(x))dif<-x[[1]]-x[[2]]
}
if(!is.null(y[1]))dif<-x-y
dif=elimna(dif)
n<-length(dif)
dif<-dif[dif!=0]  # Remove any zero values.
flag<-(dif<0)
if(!AC)temp<-binomcipv(y=flag,alpha=alpha)
if(AC)temp<-acbinomcipv(y=flag,alpha=alpha)
if(SD)temp=binomLCO(y=flag,alpha=alpha)
list(Prob_x_less_than_y=temp$phat,ci=temp$ci,n=n,N=length(flag),p.value=temp$p.value)
}