F3-mu.rank: Ranks of Data
In muStat: Prentice Rank Sum Test and McNemar Test

Description Usage Arguments Details Value Author(s) See Also Examples

Returns a vector of the (mid-) ranks of the input.

1 2	mu.rank(x, na.last = TRUE, na.rm=Inf) mu.rank.nna(x)

`x`	numeric vector. Missing values (`NA`) are allowed for `mu.rank` but not for `mu.rank.nna`
`na.last`	vector with one element. If `TRUE`, `NA`s are put last, if `FALSE`, `NA`s are put first, if `NA`, `NA`s are handled according to `na.rm`; `"keep"` is equivalent to `NA` and `na.rm` = F.
`na.rm`	logical flag, indicating if missing values (`NA`) should be removed (`TRUE`) or not (`FALSE`) in the output. If `NA`, `NA`s in `x` are not allowed and `na.last` is ignored. The default for `na.rm` is `TRUE` if `na.last = NA` and `FALSE` else.

mu.rank is faster than rank. The treatment of missing values is controlled by both na.last and na.rm.

the ranks; i.e., the i-th value is the rank of x[i]. In case of ties, average ranks is returned.

Knut M. Wittkowski kmw@rockefeller.edu, Tingting Song ttsong@gmail.com

rank

a <- c(4, 2, 5, 1, 4, NA, 6)
mu.rank(a) # default: na.last=TRUE, na.rm=FALSE
# [1] 3.5 2.0 5.0 1.0 3.5 7.0 6.0
mu.rank(a,na.last=NA) # default: na.rm=TRUE
# [1] 3.5 2.0 5.0 1.0 3.5     6.0
mu.rank(a,na.last=NA,na.rm=FALSE)
#     3.5 2.0 5.0 1.0 3.5  NA 6.0

# Spearman's rank correlation between two sets of testscores 
a <- c(4, 2, 5, 1, 4, NA, 6)
b <- c(4, 2, 5, NA, 4, 5, 6)

cor(a, b, if(is.R()) "complete.obs" else "available") 
# [1] 0.8241688
cor(a, b, if(is.R()) "pairwise.complete.obs" else "omit")
# [1] 1

cor(rank(a), rank(b))
# [1] 0.1651446
cor(mu.rank(a, na.last=NA, na.rm=FALSE),
    mu.rank(b, na.last=NA, na.rm=FALSE), 
    if(is.R()) "complete.obs" else "available")
# [1] 0.8523852 
cor(mu.rank(a, na.last=NA, na.rm=FALSE),
    mu.rank(b, na.last=NA, na.rm=FALSE), 
    if(is.R()) "pairwise.complete.obs" else "omit")
# [1] 0.9953452
cor(rank(a[!is.na(a*b)]), rank(b[!is.na(a*b)]))
# [1] 1