Exact Power [R for BE/BA]

posted by Helmut Homepage – Vienna, Austria, 2009-09-25 19:20  – Posting: # 4255
Views: 26,272

Dear all,

I made an investment and was able to compare results from R with the current version 7 of nQuery Advisor:*
+======+=====+===========+========+==========+
|  GMR |  n  |  R-code   | nQuery Advisor 7  |
+------+-----+-----------+--------+----------+
| 0.85 | 134 | 0.8017723 | 80.17  | 80.17722 |
| 0.90 |  38 | 0.8154940 | 81.54  | 81.54941 |
| 0.95 |  20 | 0.8346802 | 83.46  | 83.46802 |
| 1.00 |  16 | 0.8331980 | 83.32  | 83.32001 |
| 1.05 |  18 | 0.8001854 | 80.01  | 80.01854 |
| 1.10 |  32 | 0.8100682 | 81.00  | 81.00682 |
| 1.15 |  72 | 0.8042181 | 80.42  | 80.42180 |
| 1.20 | 294 | 0.8019247 | 80.19  | 80.19246 |
+======+=====+===========+========+==========+


nQuery requires sqrt(MSE) rather than CV as input = sqrt(ln(CV²+1)). The maximum precision is 6 decimal places. Therefore 0.1980422 rather than the value in full precision (0.1980422004353650284627675024887) like in R is used. But you see this value only if clicking the respective cell; both displayed and printed is 0.198042. nQuery gives power in percent with a maximum number of two decimal digits. That's a pity and values in the first column were the best one could get (display/print). If clicking in each cell power is given in percent to 5 decimal digits (7 significant digits); second column. It's clear that results are not rounded, but truncated. This looked weird first, but at least for power it makes sense.
Values agree nicely with the exception of a small discrepancy at GMR=1 (83.31980% vs. 83.32001%).
If I recall it correctly Dieter Hauschke wrote something about numerical problems in power estimation at T/R=1 in one of his papers.
:ponder:
I love validating software. :angry:



Cheers,
Helmut Schütz
[image]

The quality of responses received is directly proportional to the quality of the question asked. ☼
Science Quotes

Complete thread:

Activity
 Mix view
Bioequivalence and Bioavailability Forum |  Admin contact
19,478 posts in 4,133 threads, 1,333 registered users;
online 7 (0 registered, 7 guests [including 5 identified bots]).
Forum time (Europe/Vienna): 05:14 CEST

If you don’t know anything about computers,
just remember that they are machines that do exactly what you tell them
but often surprise you in the result.    Richard Dawkins

The BIOEQUIVALENCE / BIOAVAILABILITY FORUM is hosted by
BEBAC Ing. Helmut Schütz
HTML5