power_calc: Power Calculation Function

In njjms/nicks: Basic Functions for Biostatistics

Power Calculation Function

Description

Calculates power in a binomial experiment setting, given a sample size, true_prob, requirement, and significance level.

Usage

power_calc(
  sample_size,
  true_prob,
  requirement,
  requirement_type = "gt",
  alpha,
  interval_type = "ws",
  AC_type = "medium",
  prq_delta = NA
)

Arguments

sample_size	the total sample size
true_prob	the true underlying probability of success
requirement	the required underlying probability of success
requirement_type	defaults to "gt" (greater than). Can be set to "lt" (less than)
alpha	the significance level for a two-sided interval
interval_type	defaults to "ws" (wilson_score). Can be set to "cp " (Clopper-Pearson)
AC_type	specifies the acceptance criteria to be used. Defaults to "medium" (point estimate surpasses requirement). Can be set to "low" (interval includes or surpasses req), "high" (interval surpasses requirement), or "high_delta" (interval surpasses prq_delta and point estimate surpasses requirement)
prq_delta	defaults to NA. A numeric value here indicates the minimum acceptable acceptance criteria (product requirement plus/minus some delta). Only used if AC_type = "high_delta"

Value

list containing parameters of power calculation (float), df of power calculation results, and power (float)

Examples

lapply(c(.990, .985, .980),
       FUN = function(x) sapply(
         possible_sample_sizes,
         FUN = function(y)
           power_calc(sample_size = y,
                      true_prob = x,
                      requirement=desired_sensitivity,
                      alpha=significance_level*2,
                      requirement_type="gt",
                      interval_type="cp",
                      AC_type="low")$power
         )) -> sens_calculations
ggplot(
data.frame(
	n = possible_sample_sizes,
	sens_990_power=sens_calculations[[1]],
	sens_985_power=sens_calculations[[2]],
	sens_980_power=sens_calculations[[3]]
)) +
geom_line(mapping=aes(x=n, y=sens_990_power,
		  color = "true_sens_990")) +
geom_line(mapping=aes(x=n, y=sens_985_power,
		  color = "true_sens_985")) +
geom_line(mapping=aes(x=n, y=sens_980_power,
		  color = "true_sens_980")) +
scale_color_manual(name="plates",
				   values = c(true_sens_990="dodgerblue1",
							  true_sens_985="firebrick",
							  true_sens_980="forestgreen"),
				   breaks = c("true_sens_990",
							  "true_sens_985",
							  "true_sens_980")) +
labs(title="Detecting Deviation in Sensitivity from p=.995", x="n", y="Power") +
theme_bw()

njjms/nicks documentation built on May 4, 2022, 8:10 a.m.