ki.cdf: Computes CDF of KI between case profile and profile with...

In DNAprofiles: DNA Profiling Evidence Analysis

Description

Computes the Cumulative Distribution Function of a Kinship Index (KI) comparing hypotheses hyp.1 vs hyp.2 for profiles with a given relationship (hyp.true) to the case profile (e.g. "FS" for full siblings).

Usage

ki.cdf(x, hyp.1, hyp.2 = "UN", hyp.true = "UN", freqs.ki = get.freqs(x),
  freqs.true = freqs.ki, markers = {     if (missing(x))         
  intersect(names(freqs.ki), names(freqs.true))     else get.markers(x) },
  theta.ki = 0, theta.true = theta.ki, n.max = 1e+06)

Arguments

x	(optional) An integer matrix specifying a single profile.
hyp.1	A character string giving the hypothesis in the numerator of the KI. Should be one of ibdprobs, e.g. "FS" (full sibling) or "PO" (parent/offspring) or "UN" (unrelated).
hyp.2	A character string giving the hypothesis in the denominator of the KI. Should be one of ibdprobs, e.g. "FS" (full sibling) or "PO" (parent/offspring) or "UN" (unrelated). Defaults to "UN".
hyp.true	A character string specifying the true relationship between the case profile and the other profile. Should be one of ibdprobs, e.g. "FS" (full sibling) or "PO" (parent/offspring) or "UN" (unrelated). Defaults to "UN".
freqs.ki	A list specifying the allelic frequencies that are used when computing the KI.
freqs.true	(optionally) A list specifying the allelic frequencies that are used for computing the probabily distribution of the KI under hyp.true. When not provided, the function will use freqs. One might use different allelic frequencies freqs.rel when for example the case profile and relative come from some population, while KIs are computed with frequencies from another population.
markers	Character vector stating the markers for which the KI distribution is derived. By default equal to all markers of x, or, if x is missing, to the intersection of the markers of freqs.ki and freqs.true.
theta.ki	numeric value specifying the amount of background relatedness.
theta.true	numeric value specifying the amount of background relatedness.
n.max	Maximum number of events stored in memory. See dists.product.pair for details.

Examples

# for one profile, obtain the CDF of the SI,
# both for true sibs and unrelated profiles
data(freqsNLsgmplus)

# sample a profile
x <- sample.profiles(N=1,freqsNLsgmplus)

cdf.fs <- ki.cdf(x,hyp.1="FS",hyp.true="FS")
cdf.un <- ki.cdf(x,hyp.1="FS",hyp.true="UN")

# the cdf's are *functions*
cdf.fs(1)
cdf.un(1)

# we also obtain an ROC curve easily
t <- 10^(seq(from=-10,to=10,length=100)) # some thresholds
fpr <- cdf.un(t,exc.prob=TRUE)
tpr <- cdf.fs(t,exc.prob=TRUE)

plot(log10(fpr),tpr,type="l")

DNAprofiles documentation built on Jan. 15, 2017, 9:27 p.m.