###########################################################
## this code compute the power of a chisq test comparing
## the equality of several proportions (rates).
## Input:  
##   pvec   is the vector of several proportions. In most cases
##          they are not all equal (alternative). 
##        
##   nvec   the sample sizes of several groups, its success proportion
##          are been tested
###########################################################
power.chisq <- function(pvec, nvec, alpha=0.05)
{
      if( any(is.na(pvec)) || any(is.na(nvec)) ) 
         stop("there are NA in the input")
      if(length(pvec) != length(nvec))
         stop("pvec and nvec must be of same length")
      if( any(pvec >1) || any(pvec <0) ) 
         stop("pvec must all between 0 and 1")
      if( any( nvec<0 ) ) stop("negative sample size?")
      v <- length(nvec) - 1
      pbar <- sum(pvec * nvec)/sum(nvec) 
      EH0 <- pbar * nvec
      EH0 <- c(EH0, nvec - EH0)
      EHA <- pvec * nvec
      EHA <- c(EHA, nvec - EHA)
      lambda <- sum((EHA - EH0)^2/EH0)
      1 - pchisq(qchisq(1-alpha, df=v), df=v, ncp=lambda)
}