ci_from_p: CI from p value

In neatStats: Neat and Painless Statistical Reporting

CI from p value

Description

Calculates approximate confidence interval (CI) for any given difference, based on the difference value and p value, according to Altman & Bland (2011).

Usage

ci_from_p(diff, p_value, ci = 0.95)

Arguments

diff	Difference (number) around which the CI is to be calculated.
p_value	The p value for CI calculation.
ci	Numeric; confidence level for the returned CIs (.95 per default).

Value

CI limits as named numeric vector with two elements.

Note

Check the Altman & Bland (2011) paper for details! (The calculation for proportions is not implemented here.)

References

Altman, D. G., & Bland, J. M. (2011). How to obtain the confidence interval from a P value. Bmj, 343(d2090). doi: 10.1136/bmj.d2090

Examples

# Example 1
# calculate proportion difference test
proptest_stat = prop.test(x = c(49, 40), n = c(50, 50))

# calculate proportion difference
my_diff = 49/50-40/50

# calculate approximate CI
ci_from_p(my_diff, proptest_stat$p.value)

# returned CI should be very similar to the actual CI
proptest_stat$conf.int


# Example 2
# generate random data
v1 = stats::rnorm(190, 40, 60)
v2 = stats::rnorm(170, 50, 45)

# calculate t-test
ttest_stat = stats::t.test(v1, v2)

# calculate mean difference
my_diff = mean(v1) - mean(v2)

# calculate approximate CI
ci_from_p(my_diff, ttest_stat$p.value)

# returned CI should be similar to the actual CI
ttest_stat$conf.int

neatStats documentation built on Dec. 8, 2022, 1:13 a.m.