pval2index: Transform p-values in into a ranking index.

In dmontaner/mdgsa: Multi Dimensional Gene Set Analysis.

Description

After a genomic test, p-values are numerical indexes which account for certain biological characteristic. By definition p-values are bounded between zero and one, but this may not be suitable as an index. Moreover, p-values are always derived form a statistic which sign may be important. The function helps transforming the p-value and its associated statistic into a ranking index.

Usage

pval2index(pval, sign, names = NULL, log = TRUE, offset, verbose = TRUE)

Arguments

pval	a vector or matrix of p-values.
sign	a vector or matrix of signs associated to the p-values.
names	a character vector of the names of the features.
log	= TRUE
offset	value used to replace p-values equal to zero
verbose	verbose

Details

The default transformation is (-1) * log (pval) * sign (sign). When log = FALSE the transformation is (1 - pval) * sign (sign).

If sign is missing all p-values are associated wit a positive sign.

Missing values are allowed and return NA values.

An offset may be provided to replace p-values equal to zero when log = TRUE. In such way infinite values are not generated. If the offset parameter is not provided, the minimum p-value other than zero is used for the replacement. You can explicitly specify offset = 0 if you want Inf values to be returned.

By default the names of the output vector (or row names in a matrix) are those of pval or sign. If names is provided, then it is used instead.

Value

A transformed index. A vector or matrix, depending on the input parameters.

Author(s)

David Montaner [e-mail]

See Also

indexTransform

Examples

my.statistic <- rnorm (1000)
my.pvalue <- 2 * pnorm (my.statistic)
my.pvalue[my.pvalue > 1] <- 2 - my.pvalue[my.pvalue > 1]

index <- pval2index (pval = my.pvalue, sign = my.statistic)

#par (mfrow = c (1,2))
#plot (my.statistic, my.pvalue)
#plot (my.statistic, index)


## Zero p-values
p <- c (0:10)/10
p
pval2index (p)
pval2index (p, offset = 0)
pval2index (p, offset = 0.000001)

## Missing p-values
p <- c(0:10, NA)/10
p
pval2index (p)
pval2index (p, offset = 0)
pval2index (p, offset = 0.000001)
pval2index (p, log = FALSE)
pval2index (p, offset = 0, log = FALSE)

## Matrix
p <- matrix (c(0:10, NA)/10, ncol = 3)
p
pval2index (p)
pval2index (p, offset = 0)
pval2index (p, offset = 0.000001)
pval2index (p, log = FALSE)
pval2index (p, offset = 0, log = FALSE)

dmontaner/mdgsa documentation built on May 15, 2019, 9:35 a.m.