Next: findsampsize() Up: Design Macros Help File Previous: findncp()   Contents

findpower()

Usage:
findpower(means,nis,sigma2,alpha) means a REAL vector, nis a vector of
positive integers, sigma2 and alpha positive REALs; means and nis must
be the same length and alpha must be less than 1.
findpower(ncp,df1,df2,alpha), all positive scalars.



Keywords: design
                                  Usage
findpower(means,nis,sigma2,alpha) computes the power for the F-test in
the one-way anova testing the null hypothesis of no treatment
differences when the means, sample sizes, error variance, and type 1
error rate are as given.  Arguments means and nis must be vectors of the
same length (one element for each treatment).  Argument nis must be
postive integers; sigma2 must be a postive REAL, and alpha must be
between 0 and 1.

findpower(means,nis,sigma2,alpha,rcb:T) computes the power for the same
set up considered to be a randomized complete block. This requires that
all groups have the same sample size.

findpower(ncp,df1,df2,alpha) is a synonym for power2(ncp,df1,df2,alpha)

                                Examples
Here we have three treatment groups with means 2.2, 2.8, and 3.1, and
sample sizes 2, 2, and 8 respectively.  The error variance is 2 and
the type 1 error rate is .05; power is found to be .088
  Cmd> findpower(vector(2.2,2.8,3.1),vector(2,2,8),2,.05)
  (1)     0.087928

Instead suppose that we had 40 observations in each treatment, then the
power is .73
  Cmd> findpower(vector(2.2,2.8,3.1),vector(40,40,40),2,.05)
  (1)      0.72816


Gary Oehlert 2006-01-30