partialCorQUICKSTOP: QUICKSTOP significance testing for partial correlation

In bhklab/PharmacoGx: Analysis of Large-Scale Pharmacogenomic Data

QUICKSTOP significance testing for partial correlation

Description

This function will test whether the observed partial correlation is significant at a level of req_alpha, doing up to MaxIter permutations. Currently, it supports only grouping by discrete categories when calculating a partial correlation. Currenlty, only does two sided tests.

Usage

partialCorQUICKSTOP(
  pin_x,
  pin_y,
  pobsCor,
  pGroupFactor,
  pGroupSize,
  pnumGroup,
  pMaxIter,
  pn,
  preq_alpha,
  ptolerance_par,
  plog_decision_boundary,
  pseed
)

Arguments

pin_x	one of the two vectors to correlate.
pin_y	the other vector to calculate
pobsCor	the observed (partial) correlation between these varaiables
pGroupFactor	an integer vector labeling group membership, to correct for in the partial correlation. NEEDS TO BE ZERO BASED!
pGroupSize	an integer vector of size length(unique(pGroupFactor)), counting the number of members of each group (basically table(pGroupFactor)) as integer vector
pnumGroup	how many groups are there (len(pGroupSize))
pMaxIter	maximum number of iterations to do, as a REAL NUMBER
pn	length of x and y, as a REAL NUMBER
preq_alpha	the required alpha for significance
ptolerance_par	the tolerance region for quickstop. Suggested to be 1/100th of req_alpha'
plog_decision_boundary	log (base e) of 1/probability of incorrectly calling significance, as per quickstop paper (used to determine the log-odds)
pseed	A numeric vector of length 2, used to seed the internal xoroshiro128+ 1.0 random number generator. Note that currently, these values get modified per call, so pass in a copy if you wish to keep a seed for running same simulation twice

Value

a double vector of length 4, entry 1 is either 0, 1 (for TRUE/FALSE) or NA_REAL_ for significance determination NA_REAL_ is returned when the MaxIter were reached before a decision is made. Usually, this occurs when the real p value is close to, or falls within the tolerance region of (req_alpha, req_alpha+tolerance_par). Entry 2 is the current p value estimate. entry 3 is the total number of iterations performed. Entry 4 is the number of time a permuted value was larger in absolute value than the observed cor.

bhklab/PharmacoGx documentation built on April 18, 2024, 3:13 a.m.