CALF-package: Coarse Approximation Linear Function

Description Details Author(s)

Description

Forward selection linear regression greedy algorithm.

Details

The Coarse Approximation Linear Function (CALF) algorithm is a type of forward selection linear regression greedy algorithm. Nonzero weights are restricted to the values +1 and -1. The number of nonzero weights used is limited by a parameter. Samples are controls (at least 2) and cases (at least 2). A data matrix consists of a distinguished column that labels every row as either a control (0) or a case (1). Other columns (at least one) contain real number marker measurement data. Another input is a limit (positive integer) on the number of markers that can be selected for use in a linear sum. The present version uses as a score of differentiation the two-tailed, two sample unequal variance Student t-test p-value. Thus, any real-valued function applied to all samples generates values for controls and cases that are used to calculate the score. CALF selects the one marker (first in case of tie) that best distinguishes controls from cases (score is smallest p-value). CALF then checks the limit. If the number of selected markers is the limit, CALF ends. Else, CALF seeks a second marker, if any, that best improves the score of the sum function generated by adding the newly selected marker to the previous markers with weight +1 or weight -1. The process continues until the limit is reached or until no additional marker can be included in the sum to improve the score.

Author(s)

Stephanie Lane [aut, cre],
Clark Jeffries [aut],
Diana Perkins [aut], Maintainer: Stephanie Lane slane@unc.edu


stlane/calf documentation built on May 30, 2019, 5:48 p.m.