Package: Stress.dfArray
Type: Package
Title: Calculates normalization Stress and dfArray quality for a set of
        arrays
Version: 1.1
Date: 2015-10-01
Author: Douglas W. Mahoney
Maintainer: Douglas W. Mahoney <mahoney.douglas@mayo.ed>
Description: Requires a matrix form of microarray intensities (row is the feature and column is the sample) for pre- and post-normalized data. The function arrayStress calculates the normalization stress an array endures by comparing the pre- and post-normalized data. This initial pass can be used to remove arrays that have great normalization stress and we recommend that arrays with median Stress larger than 1.5 be considered as suspect. After removing arrays due to high stress, we recommend that normalization be redone on the reduced set. There are still potential issues in array quality that can not be seen by solely using the Stress metric and we have found that probe/feature behavior across the arrays is an important characteristic to consider. To this end, dfArray estimates the deviation of a specific array from the median target array of the study provided the user selects the robust approach. Another more rigours and computationally taxing approach is to use a leave one out approach much like PRESS residual diagnostics in standard linear regression modeling. As dfArray is calculated relative to feature specific standard deviation, one can easily use standard rules of thumb to judge arrays as suspect if their mediand |dfArray| is larger than some threshold (Th). In our experience, arrays with Th=1 (i.e., 50% of the arrays deviate by more than 1 standard deviation) behave quite differently than the remaining arrays. These visiual diagnostics can be seen using either dfArrayDensity or dfArrayProfile.
License: GPL(>=2)
NeedsCompilation: no
Packaged: 2015-10-01 20:46:35 UTC; sinnwell
