[rkward-devel] power analysis plugin

[Thomas Friedrichsmeier]

Hi Meik,

On Sunday 05 October 2014 16:01:27 meik michalke wrote:
> sure, why not. would someone jump in to do write the help file? ;-)

trying to write a help file sometimes helps to spot non-intuitive controls, or 
ones that could be simplified. Oh, and of course bugs...:

- For the two sample tests, when estimating the size of one of the samples, 
why not always make it the second sample (i.e. first sample size provided)?
- Syntax error for estimating sample size in different sample sizes proportion 
test.
- In some cases, you get the not-so-helpful error message:
 Error in uniroot(function(n2) eval(p.body) - power, c(2 + 1e-10, 1e+07)) : 
   f() values at end points not of opposite sign
e.g. when no finite n2 sample size is enough to satisfy the given n1 sample 
size, effect size, power, significance level. I am not sure, whether this 
applies in all cases, but where it does, it would be good to print a more 
helpful message. That might mean
  pwr.result <- try (pwr.whatever(...))
  if (class (pwr.result) == "try-error") {
     rk.print ("Power anaylsis not possible at the given specification")
     return () # Exits local(), so no need to put the rest inside an else{}.
  }
- For GLM, would it make sense to allow to specify "number of parameters to 
estimate", and "sample size (N)", instead of numerator / denominator df?
- In the generated printout() code, things might get slightly simpler, if you 
print the header earlier:
  # Prepare printout
	rk.header(pwr.result[["method"]], parameters= c (list("Target 
measure"="Whatever"), pwr.result["alternative"]))

	note <- pwr.result[["note"]]
	pwr.result[c("method", "note", "alternative")] <- NULL

	rk.results(as.data.frame (unlist (pwr.result)), titles=c ("", 
"Parameters"))
	if(!is.null(note)){
		rk.print(paste("<strong>Note:</strong> ", note))
	}
	[...]
- That said, I wonder, whether the following printout would be good enough 
(after the rk.header()):
   rk.print.literal (capture.output (print (x)))
(we could add a new function "rk.print.simple(x)" for this kind of output. It 
would probably make sense in other places, too).

Regards
Thomas
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
URL: <http://mail.kde.org/pipermail/rkward-devel/attachments/20141007/02fa1a4b/attachment.sig>