MIPHENO (version 1.2)

rm.outliers: Post-Hoc outlier removal for high throughput data

Description

Returns a dataframe with outlier groups removed. Note that each column (other than the parameter column) is treated as a separate assay. Therefore if one 'group' does not meet the criteria for inclusion in 1 assay (column value), but does for all others, only the data for the assay failing the quality control will be removed.

Usage

rm.outliers(data, parameter='FLATCODE', n=3, ...)

Arguments

data

Dataframe or Matrix of inputs for which outliers are to be removed

parameter

The parameter (given by column name) on which defines a subgroup. This is the only non-numeric column allowed.

n

the number of Median Absolute Deviations (MAD) from the global median a subgroup is allowed to be before it is considered an outlier.

...

Other parameters.

Value

rm.outliers returns a dataframe containing data with values not passing the filter converted to NA.

Details

Data should be presorted according to both identifier and parameter prior to running command. Order will be retained if this is followed. data should be a dataframe with a parameter serving as a label and the rest of the values numeric. Note that parameter should be the attribute where the most error is expected. A visual inspection using box and whisker plots may be helpful in determining the best variable to use. data is first broken down into groups based on parameter. Both the group median and the median of all groups (global median) is calculated. Groups where the absolute value of the difference between the group median and global median is greater than n * MAD(group medians) for a given attribute (column) have their values removed (ie set to NA). Data for the group is retained for columns that pass this criteria.

References

Bell SM, Burgoon LD, Last RL. MIPHENO: Data normalization for high throughput metabolite analysis. BMC Bioinformatics 2012, 13(10)

Examples

 #See the sweave document in the corresponding paper for examples