binom.conf.interval {UCS}R Documentation

Binomial Confidence Intervals

Description

Computes confidence intervals for the success probability of a binomial distribution efficiently. Unlike binom.test, this function can be applied to vectors.

Usage

binom.conf.interval(k, size, limit=c("lower","upper"),
                    conf.level=0.05, one.sided=FALSE)

Arguments

k a vector of non-negative integers. Each element represents the number of successes out of size trials, i.e. the observed value of a random variable with binomial distribution.
size a vector of positive integers. Each element represents the number of trials of a binomial distribution.
limit if "upper", the upper boundaries of the confidence intervals are returned. If "lower", the lower boundaries are returned. Note that this works both for one-sided and for two-sided confidence intervals.
conf.level the required confidence level, or rather the significance level of the corresponding binomial test (note that this behaviour differs from the built-in binom.test function). The default conf.level=0.05 stands for 95% confidence.
one.sided if TRUE, computes one-sided confidence interval (either lower or upper, depending on the value of limit). If FALSE, a two-sided confidence interval is computed (default).

Details

If one.sided=TRUE, the underlying test is one-sided (with alternative "less" or "greater", depending on the limit parameter), and the non-trivial boundary of the confidence interval is returned.

If one.sided=FALSE, the underlying test is two-sided and the requested boundary of the two-sided confidence interval is returned. For efficiency reasons, the binom.conf.interval function cheats a little and computes one-sided confidence intervals with significance level conf.level / 2.

Value

A numeric vector with the requested boundary of confidence intervals for the unknown success probabilities of binomial variables.

See Also

binom.test


[Package UCS version 0.5 Index]