MIPHENO (version 1.2)

cdf.pval: Generate Empirical pvalues from Cumulative Distribution Function

Description

Returns the pvalue for the probability of observing a response equal to the input data ( /codecdf.data or /codesample.data) or more extreme (smaller) based on an empirical distribution function (ecdf) of the cdf.data. Observations with a high pvalue (-> 1) are also rare, thus calculating 1-pvalue or 1-F will return the probability at the other end of the distribution. See ecdf for details ecdf.

Usage

cdf.pval(cdf.data, sample.data=NULL, ...)

Arguments

cdf.data

Dataframe or Matrix of inputs used to make the CDF (ie the NULL distribution or wt distribution, if known), or the sample data if a NULL distribution is unavailable.

sample.data

Optional. Dataframe or Matrix of inputs for which a pvalue is to be determined in the event that a seperate NULL distribution is used.

...

Other parameters.

Value

cdf.pval returns a dataframe containing rows= observations (input order preserved) and columns = assays. The column order may have changed to match sample.data, if provided. Values will be between 0 and 1.

Details

Data should be presorted if you are going to match it to labels (ie sample descriptors) as the labels need to be removed prior to processing. Only numeric and NA data are permitted. Columns in cdf.data and sample.data should be corresponding if using both.

Each column in the cdf.data is used for generating the CDF. For columns (assays or probes) where

References

Shannon M. Bell, Lyle D. Burgoon, Robert L. Last. MIPHENO: Data normalization for high throughput metabolite analysis. BMC Bioinformatics 2012, 13(10)

See Also

ecdf

Examples

 #See the sweave document in the corresponding paper for examples