NewmanTest: Identifying outliers with the range to standard deviation...

In lshen1/drexplorer2: Dose-response Explorer2

Description

This function implements the method described by D. Newman, Biometrika, 1939 to identify outliers.

Usage

NewmanTest(ref, obs, alpha = 0.01, recursive = FALSE)

Arguments

ref	a vector giving the reference values to estimate sigma.
obs	a vector giving the observed values where potential outlier might exist.
alpha	significance level. Only 0.01, 0.05 or 1 is allowed.
recursive	whether to recursively identify outliers. Currently not implemented.

Details

Given measurements from controls (no drug treated), we can compute the sample standard deviation (s). The range of responses from treated samples (w) can be computed for a given dose level. Assuming the controls to have the same variation as the drug treated case, the distribution of ratio statistic q=w/s can be derived and used to calculate if there is outliers in the treated responses as described by D. Newman, Biometrika, 1939.

Note that this function works for a single dose level. When multiple dose levels exist, one need to repeatedly call this function to identify outliers at each dose level or use the flagOutliers() function which is just a wrapper.

Value

indicator a logical vector specifying if the corresponding point is flagged as outlier.

References

Newman, D. (1939). The distribution of range in samples from a normal population, expressed in terms of an independent estimate of standard deviation. Biometrika, 31(1/2), 20-30.http://www.jstor.org/stable/2334973

Examples

set.seed(1)
x <- rnorm(10, 0, 1)
y <- c(rnorm(5, 0, 1), rnorm(1, 0, 1) + 4)
# the last observation in y is an outlier
NewmanTest(x, y, alpha = 0.05)

lshen1/drexplorer2 documentation built on June 2, 2020, 9:27 p.m.