Log in
Register|
Helmut ★★★   Vienna, Austria, 2011-02-07 20:05 (4820 d 21:03 ago) Posting: # 6574 Views: 5,962 |
|
|
Dear all, at the recent workshop in Mumbai I struggled (again) to deliver the message that a posteriori power does not make sense. In the aftermath of this post on Potvin et al.'s paper I found a sentence soothing my soul: We stop here and conclude BE, irrespective of the fact that we have not yet achieved the desired power of 80% (power = 66.3%). — Dif-tor heh smusma 🖖🏼 Довге життя Україна! ![[image]](https://static.bebac.at/pics/Blue_and_yellow_ribbon_UA.png) Helmut Schütz ![[image]](https://static.bebac.at/img/CC by.png) The quality of responses received is directly proportional to the quality of the question asked. 🚮 Science Quotes |
|
d_labes ★★★ Berlin, Germany, 2011-02-08 10:18 (4820 d 06:49 ago) @ Helmut Posting: # 6580 Views: 5,035 |
|
|
Dear Helmut, I recently had a question from an author of a paper in a medical journal. He had stated "A Student's t-test of ...  give a non-significant result."Statement of a reviewer to that sentence: "A power analysis is expected." [sic !] — Regards, Detlew |
|
Helmut ★★★ Vienna, Austria, 2011-02-08 12:27 (4820 d 04:41 ago) @ d_labes Posting: # 6588 Views: 5,109 |
|
|
Dear D. Labes! ❝ [...] "A power analysis is expected." [sic !] Hurrah! Maybe the reviewer should have been guided to CONSORT, especially the download section. The 2010 guideline was published in leading journals - maybe he/she doesn't read? Quote from Item 7 (Sample size): There is little merit in a post hoc calculation of statistical power using the results of a trial; the power is then appropriately indicated by confidence intervals. See also this nice discussion in the British Medical Journal after CONSORT 2001 was published. — Dif-tor heh smusma 🖖🏼 Довге життя Україна! Helmut Schütz The quality of responses received is directly proportional to the quality of the question asked. 🚮 Science Quotes |
|
earlybird ☆ 2011-02-08 12:20 (4820 d 04:47 ago) @ Helmut Posting: # 6586 Views: 5,123 |
|
|
Dear HS, you should take into considerations a statistic course for regulators  . As long as the regulators ask for the a-posteriori-power, you should be able to calculate it  .Kind reagards, earlybird |
|
Helmut ★★★ Vienna, Austria, 2011-02-08 13:22 (4820 d 03:46 ago) @ earlybird Posting: # 6589 Views: 5,160 |
|
|
Dear earlybird! ❝ you should take into considerations a statistic course for regulators Well, I was asked by regulators from Austria and Saudi Arabia a good while ago. They were shocked by the fee.  I performed a workshop in Turkey; was anybody of you asked for power (in other words, did I succeed)? ❝ As long as the regulators ask for the a-posteriori-power, … Do they still? Which ones? I know that this was a hobby of regulators from Greece more than a decade ago. The last request I have seen came form Malaysia. They were asking for ‘Two One Sided t-Tests Total Probability’, ‘Anderson-Hauck Procedure p-value’, ‘Power of ANOVA at 20%’. By coincidence (?) these are the last lines of WinNonlin’s standard output. ❝ … you should be able to calculate it ⇒ PowerTOST! — Dif-tor heh smusma 🖖🏼 Довге життя Україна! Helmut Schütz The quality of responses received is directly proportional to the quality of the question asked. 🚮 Science Quotes |
|
Dr_Dan ★★ Germany, 2011-02-08 15:31 (4820 d 01:37 ago) @ Helmut Posting: # 6591 Views: 4,989 |
|
|
Dear Helmut dear all some time ago I had to answer the same question but since I am not a statistician I pointed out the following which is in my opinion the assessor's concern: The risk for the patient that the study might show a false positive result is always 5%. The power only describes the risk for the sponsor to fail. Right? LG Dan — Kind regards and have a nice day Dr_Dan |
|
Helmut ★★★ Vienna, Austria, 2011-02-08 15:34 (4820 d 01:34 ago) @ Dr_Dan Posting: # 6592 Views: 5,039 |
|
|
Dear Dan! ❝ The risk for the patient that the study might show a false positive result is always 5%. The power only describes the risk for the sponsor to fail. Right? Right. — Dif-tor heh smusma 🖖🏼 Довге життя Україна! Helmut Schütz The quality of responses received is directly proportional to the quality of the question asked. 🚮 Science Quotes |
|
d_labes ★★★ Berlin, Germany, 2011-02-08 16:17 (4820 d 00:51 ago) @ Helmut Posting: # 6595 Views: 5,024 |
|
|
Dear Dan, dear Helmut! ❝ ❝ The risk for the patient that the study might show a false positive result is always 5%. Right? ❝ ❝ Right. For answering regulatory concerns it is the best answer I know  .But let us keep in mind the so called "test size" of the statistical test method used. In many cases the test method assures an size <= 0.05, thus the test may become conservative under circumstances. Then the patient's risk is at maximum 5% and otherwise <5%. In case of "liberal" test methods the patients risk may become >5%. Such test methods should of course not used. But sometimes you have no choice ... Remember f.i. Potvin et.al. methods for evaluating 2-stage designs which have been chosen under the assumption that 5.2% alpha inflation are allowed, considered as negligible. Or remember the FDA scaled ABE approach according to the progesterone guidance which leads to an alpha inflation up to 6% - 7%. See here. — Regards, Detlew |
Ing. Helmut Schütz