KEEP ROWS WITH THE MAXIMUM STATISTIC

A common task in expression analysis is to collapse multiple features that are mapped to the same gene by some statistic. This function does this job by keeping the matrix row (normally features) with the higheest statistic specified by the user.

Usage

keepMaxStatRow(
  matrix,
  keys,
  keepNArows = TRUE,
  stat = function(x) mean(x, na.rm = TRUE),
  levels = c("rownames", "attribute", "discard"),
  ...
)

Arguments

matrix

A numeric matrix

keys

A vector of character giving the keys the rows are mapped to. A common scenario is that each row represents one probeset, while the vector keys give the genes that the probesets are mapped to. Thus keys can be redundant, namely multiple probesets can map to the same gene.

keepNArows

Logical, whether rows with NA as their keys should be kept (TRUE) or should be discarded (FALSE)

stat

The function to calculate the univariate statistic. By default the NA-robust mean is used.

levels

How should the information of the levels of keys, e.g. unique keys, be kept. dicard will discard this information, rownames will make the unique keys (potentially with NAs) as row names of the output matrix, and attribute will append an attribute named levels to the output matrix.

...

Other parameters passed to the stat function

Value

A numeric matrix with rows mapped to unique keys, selected by the maximum statistics. See examples below

Details

isMaxStatRow returns a logical vector, with rows with maximal statistics each key as TRUE and otherwise as FALSE. keepMaxStatRowInd returns the integer indices of such rows. Finally keepMaxStatRow returns the resulting matrices.

For use see examples

Author

Jitao David Zhang <jitao_david.zhang@roche.com>

Examples

myFun1 <- function(x) mean(x, na.rm=TRUE)
myFun2 <- function(x) sd(x, na.rm=TRUE)
mat1 <- matrix(c(1,3,4,-5,
                 0,1,2,3,
                 7,9,5,3,
                 0,1,4,3), ncol=4, byrow=TRUE)
keys1 <- c("A", "B", "A", "B")

isMaxStatRow(mat1, keys1, stat=myFun1)
#> [1] FALSE FALSE  TRUE  TRUE
isMaxStatRow(mat1, keys1, stat=myFun2)
#> [1]  TRUE FALSE FALSE  TRUE

keepMaxStatRowInd(mat1, keys1, stat=myFun1)
#> [1] 3 4
keepMaxStatRowInd(mat1, keys1, stat=myFun2)
#> [1] 1 4

keepMaxStatRow(mat1, keys1, stat=myFun1)
#>   [,1] [,2] [,3] [,4]
#> A    7    9    5    3
#> B    0    1    4    3
keepMaxStatRow(mat1, keys1, stat="myFun2")
#> Error in get(stat, envir = parent.frame()): object 'myFun2' not found
keepMaxStatRow(mat1, keys1, stat="myFun2", levels="discard")
#> Error in get(stat, envir = parent.frame()): object 'myFun2' not found
keepMaxStatRow(mat1, keys1, stat="myFun2", levels="attribute")
#> Error in get(stat, envir = parent.frame()): object 'myFun2' not found

mat2 <- matrix(c(1,3,4,5,
                 0,1,2,3,
                 7,9,5,3,
                 0,1,4,3,
                 4,0,-1,3.1,
                 9,4,-3,2,
                 8,9,1,2,
                 0.1,0.2,0.5,NA,
                 NA, 4, 3,NA), ncol=4, byrow=TRUE,
               dimnames=list(LETTERS[1:9], NULL))
keys2 <- c("A", "B", "A", "B", NA, NA, "C", "A", "D")

isMaxStatRow(mat2, keys2, keepNArows=FALSE, stat=myFun1)
#> [1] FALSE FALSE  TRUE  TRUE FALSE FALSE  TRUE FALSE  TRUE
keepMaxStatRowInd(mat2, keys2, keepNArows=FALSE, stat=myFun1)
#> [1] 3 4 7 9

keepMaxStatRow(mat2, keys2, keepNArows=FALSE, stat=myFun1)
#>   [,1] [,2] [,3] [,4]
#> A    7    9    5    3
#> B    0    1    4    3
#> C    8    9    1    2
#> D   NA    4    3   NA
keepMaxStatRow(mat2, keys2, keepNArows=TRUE, stat=myFun1)
#>      [,1] [,2] [,3] [,4]
#> A       7    9    5  3.0
#> B       0    1    4  3.0
#> <NA>    4    0   -1  3.1
#> <NA>    9    4   -3  2.0
#> C       8    9    1  2.0
#> D      NA    4    3   NA
keepMaxStatRow(mat2, keys2, keepNArows=TRUE, stat=myFun1, levels="discard")
#>   [,1] [,2] [,3] [,4]
#> C    7    9    5  3.0
#> D    0    1    4  3.0
#> E    4    0   -1  3.1
#> F    9    4   -3  2.0
#> G    8    9    1  2.0
#> I   NA    4    3   NA
keepMaxStatRow(mat2, keys2, keepNArows=TRUE, stat=myFun1, levels="attribute")
#>   [,1] [,2] [,3] [,4]
#> C    7    9    5  3.0
#> D    0    1    4  3.0
#> E    4    0   -1  3.1
#> F    9    4   -3  2.0
#> G    8    9    1  2.0
#> I   NA    4    3   NA
#> attr(,"levels")
#> [1] "A" "B" NA  NA  "C" "D"