power_mcnemar_test: Power Calculations for Exact and Asymptotic McNemar Test in a 2 by 2 table

Description

Compute power of test, or determine parameters to obtain target power for matched case-control studies.

Usage

power_mcnemar_test(
  n = NULL,
  paid = NULL,
  psi = NULL,
  sig.level = 0.05,
  power = NULL,
  alternative = c("two.sided", "one.sided"),
  method = c("normal", "exact", "cond.exact")
)

Arguments

Number of observations (number of pairs)

paid

The probability that a case patient is not exposed and that the corresponding control patient was exposed (specifying p_12 in the 2 x 2 table). It is assumed that this is the _smaller_ of the two discordant probabilities.

psi

The relative probability that a control patient is not exposed and that the corresponding case patient was exposed compared to the probability that a case patient is not exposed and that the corresponding control patient was exposed (i.e., p_21 / p_12 in the 2x2 table). Also called the discordant proportion ratio. psi must be larger than or equal to 1 since paid was the smaller of the two discordant probabilities.

sig.level

Significance level (Type I error probability)

power

Power of test (1 minus Type II error probability)

alternative

One- or two-sided test

method

Power calculations based on exact or asymptotic test. The default (normal) corresponds to an approximative test, "exact" is the unconditional exact test, while "cond.exact" is a conditional exact test (given fixed n). The "exact" method is very slow for large values of n so it is most useful for fixed (and moderately-sized) n.

Value

Object of class power.htest, a list of the arguments (including the computed one) augmented with method and note elements.

References

Duffy, S (1984). Asymptotic and Exact Power for the McNemar Test and its Analogue with R Controls per Case

Fagerland MW, Lydersen S, Laake P. (2013) The McNemar test for binary matched-pairs data: mid-p and asymptotic are better than exact conditional. BMC Medical Research Methodology.

Examples

Run this code

# NOT RUN {
# Assume that pi_12 is 0.125 and we wish to detect an OR of 2.
# This implies that pi_12=0.25, and with alpha=0.05, and a power of 90% you get
power_mcnemar_test(n=NULL, paid=.125, psi=2, power=.9)

power_mcnemar_test(n=NULL, paid=.1, psi=2, power=.8, method="normal")
power_mcnemar_test(n=NULL, paid=.1, psi=2, power=.8)



# }