Bioequivalence and Bioavailability Forum • Some minor comments

Some minor comments [Power / Sample Size]

posted by Ben – Germany, 2011-11-23 22:12 (5322 d 09:01 ago) – Posting: # 7704
Views: 10,262

Thank you very much for your effort and help!

❝ ... I strongly argue that the differences are more probably due to the different number of degrees of freedom, namely df=n-1 in the so-called 'paired' design. You (having access to nQuery) can test it with my code suggestions below.

You're absolutely right! I didn't check very thoroughly but so far it seems that exactly these df are being used (this indeed makes sense as the CI is for difference in "paired means")

❝ Here some minor code suggestions: ...

You put in some nice features (starting value of n and the step size), awesome. One suggestion/question from myself: should it be n <- step*ceiling(n/step) instead of n <- step*trunc(n/step)? Otherwise we will get an error using e.g. w=0.3, sigma=0.2, design="2x2".

Also, one could think of including n=NULL or something like that as input parameter and then return the precision if a sample size is given, that is,


if (is.null(n)) {

...

} else {

   df <- eval(dfe)

   if (df < 1) {

   stop("n is too small!", call. = FALSE)

   }

   w <- sigma*sqrt(qchisq(1-gamma,df)/df*bk/n*qf(1-alpha,1,df))

   return(w)

}

The values of w coincide with those of nQuery (using "paired").

Best regards

Complete thread:

Trials to a given precision; Sample size Ben 2011-11-14 19:16
- Sample size for a given precision of CI d_labes 2011-11-16 09:40
  - Sample size for a given precision of CI Ben 2011-11-18 16:57
    - Sample size for a given precision of CI Ben 2011-11-21 19:13
      - Some minor comments d_labes 2011-11-22 15:37
        
        Some minor commentsBen 2011-11-23 21:12
        
        Some minor comments Ben 2011-11-25 17:28
        
        Using undocumented functions - PowerTOST v0.9-0 d_labes 2011-12-15 09:28
        
        Using undocumented functions - PowerTOST v0.9-0 Ben 2011-12-16 19:06