Bioequivalence and Bioavailability Forum

Exact power and sample size, Part II [🇷 for BE/BA]

posted by d_labes  – Berlin, Germany, 2009-11-23 15:38 (6042 d 06:49 ago) – Posting: # 4375
Views: 14,687

Dear All!

Here comes part II of the story:

# a bunch of functions to get the design characteristics
known.designs<-function()
{
# Note: the df for the replicate designs are those without carry-over
# n is the total number in case of cross-over design,
# n is number per group in case of parallel group design
# df2 = degrees of freedom for robust analysis (aka Senn's basic estimator)
# bk is the so-called design constant
  des <- ("no  design    df      df2    steps  bk
           0  parallel 2*(n-1) 2*(n-1)  1      2
           1  2x2      n-2     n-2      2      2
           2  3x3      2*n-3   n-3      3      2
           3  4x4      3*n-5   n-4      4      2
           4  2x2x3    2*n-3   n-2      2      1.5
           5  2x2x4    3*n-4   n-2      2      1
           6  2x4x4    3*n-4   n-4      4      1")
# eventually it would be better to have steps=6 in case of 3x3 (6 seq. design)
# the df2 have to be checked! Not used so far. TODO !
   
  des2 <- textConnection(des)
  designs <- read.table(des2, header=TRUE, sep="", strip.white=TRUE, as.is=TRUE)           
  close(des2)   # without this close() warnings are generated
 
  # names for nicer output of design
  designs$name[designs$no==0] <- "2 parallel groups"
  designs$name[designs$no %in% c(1,2,3)] <-
      paste(designs$design[designs$no %in% c(1,2,3)],"crossover")
  designs$name[designs$no %in% c(4,5,6)] <-
      paste(designs$design[designs$no %in% c(4,5,6)],"replicate crossover")

  return(designs)

}
#--- return no of design ---
# design: a character string describing the design
.design.no<-function(design)
{
  #take the first word if more then one
  desi <- unlist(strsplit(tolower(design)," "))[1]
  des <- known.designs()
  i   <- match(desi, des$design)
  if (!is.na(i)) return(des$no[i]) else return(NA)
}
#--- return design constant ---
.design.bk<-function(design.no)
{
  des <- known.designs()
  return(des$bk[des$no==design.no])
}
#--- return design degrees of freedom ---
# as expression
.design.df<-function(design.no, nvar="n")
{
  des <- known.designs()
  df  <- (des$df[des$no==design.no])
  #replace with calling variable name
  df  <- gsub('n',nvar,df)
  e   <- parse(text=df, srcfile=NULL)
  return(as.expression(e))
}
#--- return step for sample size search ---
.design.step<-function(design.no)
{
  des <- known.designs()
  return (des$steps[des$no==design.no])
}
#--- return step for sample size search ---
.design.name<-function(design.no)
{
  des <- known.designs()
  return (des$name[des$no==design.no])
}

Continuation follows.

Again: Use it on your own risk! The author does not take any responsibility for
damage of your computer, your career or of the world .

Complete thread:

• Exact power and sample size, Part I d_labes 2009-11-23 14:32 
  • Exact power and sample size, Part IId_labes 2009-11-23 14:38
    • Exact power and sample size, Part III d_labes 2009-11-23 14:50
      • Exact power and sample size, Part IV d_labes 2009-11-23 14:57
        • Backslashes in R-code Helmut 2009-11-23 15:13
          • Backslashes eater d_labes 2009-11-23 15:19
            • Backslashes eater Helmut 2009-11-23 15:28