Bioequivalence and Bioavailability Forum • R-code shortening

R-code shortening [Power / Sample Size]

posted by d_labes – Berlin, Germany, 2014-08-13 18:32 (3926 d 09:59 ago) – Posting: # 13375
Views: 19,820

Dear Helmut,

here comes my programming gem :cool:

for the sensitivity with respect to sample size: Replace all the code for the different designs by

  ####################################

  # min. n for minimum accept. power #

  # workaround, since uniroot() does #

  # not accept two vectors as limits #

  ####################################

  Ns <- seq(n.est, 12, by=-1) # don't drop below 12 subjects

  j <- 0

  pwrN <- pwr.est

  n.min <- NULL; pBEn <- NULL

  # get # of sequences f.i. via this clumsy construct

  seqs <- known.designs()[known.designs()$design==des,"steps"]


  n    <- vector("numeric", length=seqs)

  ni   <- 1:seqs

  # may it be that j grows greater than length(Ns)?

  # paranoia

  nNs  <- length(Ns)

  while(pwrN >= pwr.min & j<nNs){

    j <- j+1

    # distribute total Ns to the sequence groups

    # n[-seqs] is n[1:(seqs-1)]

    n[-seqs] <- diff(floor(Ns[j]*ni/seqs))

    n[seqs]  <- Ns[j] -sum(n[-seqs])

    # can use allways power2.TOST() if PowerTOST V1.1-13 is used

    # else an if() construct is necessary switching to power.TOST()

    # in case of des=="paired"

    pwrN <- power2.TOST(CV=CV, n=n, theta0=GMR, design=des)

    if(pwrN >= pwr.min) {

      # stepwise growing vectors is R-Inferno Circle 2

      # but here we dont know the dimension in advance. Right?

      n.min <- c(n.min, sum(n))

      pBEn  <- c(pBEn, 100*pwrN)

    } else {

      break

    }

  }

  # plot code follows

  ...

—
Regards,

Detlew

Complete thread:

Deviating from assumptions Helmut 2014-08-08 15:44 [Power / Sample Size]
- Some Nitpicking d_labes 2014-08-12 09:12
  - Some Nitpicking Helmut 2014-08-12 10:49
  - R-code for all ABE designs Helmut 2014-08-12 17:21
    - Sensitivity analysis for all ABE designs d_labes 2014-08-13 09:30
      - Sensitivity analysis for all ABE designs Helmut 2014-08-13 14:49
    - R-code shorteningd_labes 2014-08-13 16:32
      - Suggestions / Sneak Preview Helmut 2014-08-13 16:42
        
        Suggestions / Sneak Preview d_labes 2014-08-15 09:01
        
        Suggestions / Sneak Preview Helmut 2014-08-15 12:02
        
        Mehl returned! d_labes 2014-08-15 11:27
        
        Mehl returned! Helmut 2014-08-15 11:42