plotPower: Make a power curve

In dkahle/bayesRates: Two-Sample Tests and Sample Size Determination from a Bayesian Perspective

Description

Make a power curve

Usage

plotPower(ts, a1, b1, a2, b2, a = a1, b = b1, pi0 = 0.5, pi1 = 1 -
  pi0, c = 1, family = c("binomial", "poisson"), ylim = 0:1)

Arguments

ts	vector specifying x-range
a1	alpha, the hyperparameter of the gamma distribution of the first poisson rate
b1	beta, the hyperparameter of the gamma distribution of the first poisson rate
a2	alpha, the hyperparameter of the gamma distribution of the second poisson rate
b2	beta, the hyperparameter of the gamma distribution of the second poisson rate
a	alpha, the hyperparameter of the gamma distribution under the null
b	beta, the hyperparameter of the gamma distribution under the null
pi0	the prior probability of the null hypothesis
pi1	the prior probability of the alternative hypothesis
c	relative loss constant (loss due to type II error divided by loss due to type I error)
family	"binomial" or "poisson", depending on test
ylim	y limits

Examples

## Not run: 

a1 <- 2; b1 <- 6
a2 <- 3; b2 <- 3
plotPower(ts = c(0,250), a1, b1, a2, b2)
plotPower(ts = c(0,1000), a1, b1, a2, b2)
plotPower(ts = c(0,100), a1, b1, a2, b2)
plotPower(ts = 0:300, a1, b1, a2, b2, method = 'exact')
# this last plot shows that the approximation is not quite exact for small values
# but decent for larger ones

# styling, all ggplot2 styling works
plotPower(t = c(0,1000), a1, b1, a2, b2) +
  theme_bw() + opts(title = 'My Power Plot')



a1 <- 4; b1 <- 4
a2 <- 8; b2 <- 4
plotPower(t = c(0,200), a1, b1, a2, b2, family = "poisson") +
  geom_hline(yintercept = .80)







a1 <- 4; b1 <- 2
a2 <- 6; b2 <- 2
plotPower(t = c(0,1000), a1, b1, a2, b2)
plotPower(t = c(0,25), a1, b1, a2, b2)
plotPower(t = 0:25, a1, b1, a2, b2, method = 'exact')




a1 <- c(2,4); b1 <- c(6,2);
a2 <- c(3,6); b2 <- c(3,2);
plotPower(t = c(0,1000), a1, b1, a2, b2)

library(ggplot2)
last_plot() + theme_bw()




 
## End(Not run)

dkahle/bayesRates documentation built on May 15, 2019, 9:07 a.m.