MCELESTINO
Junior

Mexico,
2019-01-03 22:05
(edited by mittyri on 2019-01-05 02:33)

Posting: # 19725
Views: 562
 

 power_TOST VS statistical power of the ANOVA [Power / Sample Size]

In the results it is also evident that the post-hoc statistical power (power_TOST) is less than 80%, this is because the estimated point is greater than 105%, considering that the sample size was estimated expecting a difference less than / equal to 5%. However, the statistical power of the ANOVA (Power_80_20) supports that the sample size used in the study was adequate, since the values ​​are greater than 90%.

This is true?


Please follow the Forum’s Policy[Mittyri]
Please don’t shout[Helmut]
mittyri
Senior

Russia,
2019-01-04 00:24

@ MCELESTINO
Posting: # 19726
Views: 474
 

 Power_80_20

Hi,

» In the results it is also evident that the post-hoc statistical power (power_TOST) is less than 80%, this is because the estimated point is greater than 105%, considering that the sample size was estimated expecting a difference less than / equal to 5%.

There are 3 basic factors in power calc: sample size, CV and PE. In post-hoc analysis sample size is fixed, but do not forget about CV! So the conclusion is valid if CV is equal or less than expected

» However, the statistical power of the ANOVA (Power_80_20) supports that the sample size used in the study was adequate, since the values ​​are greater than 90%.

:-D I think it should go to the trash bin. That's a conclusion based on WNL output, right?
See this old thread
That 'magick power' is not a power of ANOVA (whenever it could be) but the result of equality test using delta = 20%. That test has much more power than PowerTOST, thus, the value is higher than the previous one.
You cannot assess the sample size using that test.

Kind regards,
Mittyri
MCELESTINO
Junior

Mexico,
2019-01-04 19:18
(edited by mittyri on 2019-01-05 02:36)

@ mittyri
Posting: # 19729
Views: 425
 

 Power_80_20

Hi,
» There are 3 basic factors in power calc: sample size, CV and PE. In post-hoc analysis sample size is fixed, but do not forget about CV! So the conclusion is valid if CV is equal or less than expected

If CV is higher than expected,
What is the conclusion?

Thanks
Regards



Edit: Standard quotes restored. [Mittyri]
Please don’t shout. [Helmut]
ElMaestro
Hero

Denmark,
2019-01-04 21:39

@ MCELESTINO
Posting: # 19730
Views: 417
 

 Power_80_20

Hello McElestino,

» If CV is higher than expected,
» What is the conclusion?

That is your conclusion. Or parts of it.

if (3) 4

x=c("Foo", "Bar")
b=data.frame(x)
typeof(b[,1]) ##aha, integer?
b[,1]+1 ##then let me add 1



Best regards,
ElMaestro

"(...) targeted cancer therapies will benefit fewer than 2 percent of the cancer patients they’re aimed at. That reality is often lost on consumers, who are being fed a steady diet of winning anecdotes about miracle cures." New York Times (ed.), June 9, 2018.
Activity
 Thread view
Bioequivalence and Bioavailability Forum |  Admin contact
19,268 posts in 4,098 threads, 1,314 registered users;
online 16 (1 registered, 15 guests [including 9 identified bots]).

That which is not controversial
is of no particular interest.    Johann Wolfgang von Goethe

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