statpsych for Sample-Size Planning

One great tool for sample-size planning is the statpsych package by Doug Bonett.

There are loads and loads of goodies in statpsych, and the documentation is excellent and authoritative.

Here, for example, is how you can find the sample-size needed to obtain a desired level of power for a predicted effect in a 2-group design

# If you don't have statpsych installed, install it!
if (!require("statpsych")) install.packages("statpsych")

Loading required package: statpsych

# size.test.mean2 gives power for the mean difference in a 2-group design
# parameters are alpha level, desired power, average within-group variance, effect size, and ratio of group1 to group2 sample sizes
# Full documentation is at: https://dgbonett.github.io/statpsych/reference/size.test.mean2.html
# In this example, we have 
#  * an alpha of .05, 
#  * desired power of .95, 
#  * expected average variance of 100 (sd_avg = 10)
#  * an predicted effect size of 10
#  * and equal sample-sizes in both groups (ratio of 1 between n1 and n2)
# The output shows that we need n = 27 per group.
statpsych::size.test.mean2(.05, .95, 100, 10, 1) 

 n1 n2
 27 27

# We can easiy investigate other effect sizes, power levels, etc.  For example:
#  Here we explore effect sizes of 5, 10, and 20 while keeping
#  all other parameters constant
for(e_size in c(5, 10, 20)){
  print(
    paste(
      "For effect size of", e_size,
      "need sample size of", 
      paste(statpsych::size.test.mean2(.05, .95, 100, e_size, 1), collapse = ", ")
    )
  )
  
}

[1] "For effect size of 5 need sample size of 105, 105"
[1] "For effect size of 10 need sample size of 27, 27"
[1] "For effect size of 20 need sample size of 8, 8"