ci.indep.interaction: Confidence interval for the interaction of independent...

In m-Py/propint: Testing for interactions in proportions

Description

This function tests for an interaction effect in independent proportions: it tests the whether the difference between two proportions is larger than the difference between two different proportions using a confidence interval (i.e., the parameter of interest is d = (p1 - p2) - (p3 - p4)). This corresponds closely to a 2x2 ANOVA when outcomes are dichotomous rather than continuous. See Newcombe (2001).

Usage

ci.indep.interaction(ci = NULL, a1 = NULL, m1 = NULL, b1 = NULL,
  n1 = NULL, a2 = NULL, m2 = NULL, b2 = NULL, n2 = NULL, p1 = NULL,
  l1 = NULL, u1 = NULL, p2 = NULL, l2 = NULL, u2 = NULL, p3 = NULL,
  l3 = NULL, u3 = NULL, p4 = NULL, l4 = NULL, u4 = NULL)

Arguments

ci	The confidence level - pass as a natural number (e.g. 95 for the 95% confidence interval). Only pass this argument when passing absolute frequencies (a, m , b, n); do not combine with proportions and confidence interval boundaries (in this case the confidence interval is implied in the boundaries from the passed intervals).
a1	Frequency of events in proportion A1
m1	Total frequency in proportion A1
b1	Frequency of events in proportion B1
n1	Total frequency in proportion B1
a2	Frequency of events in proportion A2
m2	Total frequency in proportion A2
b2	Frequency of events in proportion B2
n2	Total frequency in proportion B2
p1	Proportion 1 - a value between 0 and 1. Pass only if absolute frequencies (a, m , b, n) are not given.
l1	The lower bound of the confidence interval of p1.
u1	The upper bound of the confidence interval of p1.
p2	Proportion 2 - a value between 0 and 1. Pass only if absolute frequencies (a, m , b, n) are not given.
l2	The lower bound of the confidence interval of p2.
u2	The upper bound of the confidence interval of p2.
p3	Proportion 3 - a value between 0 and 1. Pass only if absolute frequencies (a, m , b, n) are not given.
l3	The lower bound of the confidence interval of p3.
u3	The upper bound of the confidence interval of p3.
p4	Proportion 4 - a value between 0 and 1. Pass only if absolute frequencies (a, m , b, n) are not given.
l4	The lower bound of the confidence interval of p4.
u4	The upper bound of the confidence interval of p4.

Value

A list containing the confidence interval boudaries.

d	The difference between differences (d = (a1/m1 - b1/n1) - (a2/m2 - b2/n2) = (p1 - p2) - (p3 - p4)
l	The lower bound of the confidence interval
u	The upper bound of the confidence interval

Author(s)

Martin Papenberg [e-mail]

References

Newcombe, R. G. (2001). Estimating the difference between differences: measurement of additive scale interaction for proportions. Statistics in medicine, 20(19), 2885-2893.

Examples

# Example from From Newcombe (2001):

ci.indep.interaction(ci=95, a1=17, m1=65, b1=17, n1=75, a2=18, m2=72,
b2=16, n2=65)

# Alternative usage:
p1 <- ci.one.prop(95, 17, 65)
p2 <- ci.one.prop(95, 17, 75)
p3 <- ci.one.prop(95, 18, 72)
p4 <- ci.one.prop(95, 16, 65) # it is possible to use other CI boundaries
ci.indep.interaction(p1=p1$p, p2=p2$p, p3=p3$p, p4=p4$p, l1=p1$l, l2=p2$l, l3=p3$l, l4=p4$l,
                     u1=p1$u, u2=p2$u, u3=p3$u, u4=p4$u)

m-Py/propint documentation built on July 26, 2019, 8:47 a.m.