PowerTOST and Sample size estimation for R-SABE [RSABE / ABEL]
Dear All!
I like the giraffe finally understood the power of scaled approach and I am shocked by the required sample size...
Does it mean that we DO NOT need more than 40 subjects for ANY HVD with CV 30-60% bearing in mind just GMR about 90-110% in four period replicate study?
In order to compare sampleN.TOST unscaled results with sampleN.scABEL I calculated several points for CV in the range 0.3-0.6, assuming thetha0=0.9, design="2x2x4".
The results of sampleN.scABEL naturally coincides with those of Tóthfalusi et Endrényi (in brackets):
As far as I know sampleN.scABEL simulates studies to get the target power so it would be erroneously just to calculate sample size using sampleN.TOST and scale theta1 and theta2, wouldn't it?
The graph for sampleN.scABEL results has a natural minimum at CV=50% (as far as I understand it is because of the EMA limit on scale).
The difference between unscaled and scaled results is drastic. It doesn't fit in my head because there exist replicate studies completed with more than 50 subjects... It turns that it was redundantly?
![[image]](img/uploaded/image316.png)
I like the giraffe finally understood the power of scaled approach and I am shocked by the required sample size...
Does it mean that we DO NOT need more than 40 subjects for ANY HVD with CV 30-60% bearing in mind just GMR about 90-110% in four period replicate study?
In order to compare sampleN.TOST unscaled results with sampleN.scABEL I calculated several points for CV in the range 0.3-0.6, assuming thetha0=0.9, design="2x2x4".
The results of sampleN.scABEL naturally coincides with those of Tóthfalusi et Endrényi (in brackets):
CV 0.3 34 (35)
CV 0.35 30 (34)
CV 0.4 30 (31)
CV 0.45 28 (29)
CV 0.5 28 (28)
CV 0.55 30 (30)
CV 0.6 32 (32)
As far as I know sampleN.scABEL simulates studies to get the target power so it would be erroneously just to calculate sample size using sampleN.TOST and scale theta1 and theta2, wouldn't it?
The graph for sampleN.scABEL results has a natural minimum at CV=50% (as far as I understand it is because of the EMA limit on scale).
The difference between unscaled and scaled results is drastic. It doesn't fit in my head because there exist replicate studies completed with more than 50 subjects... It turns that it was redundantly?
![[image]](img/uploaded/image316.png)
Complete thread:
- Sample size calculation for R-SABE Lucas 2014-02-07 11:41 [RSABE / ABEL]
- EMA’s ABEL Helmut 2014-02-07 14:21
- EMA’s ABEL Lucas 2014-02-07 16:49
- Carry-over Helmut 2014-02-07 17:12
- Carry-over Lucas 2014-02-07 17:42
- Carry-over Shuanghe 2014-02-10 17:11
- Carry-over Lucas 2014-02-07 17:42
- Carry-over Helmut 2014-02-07 17:12
- EMA’s ABEL Lucas 2014-02-07 16:49
- PowerTOST and Sample size estimation for R-SABE d_labes 2014-02-07 14:37
- PowerTOST and Sample size estimation for R-SABE kumarnaidu 2014-04-18 13:38
- PowerTOST and Sample size estimation for R-SABE Dr_Dan 2014-04-18 19:29
- PowerTOST and Sample size estimation for R-SABE kumarnaidu 2014-04-19 06:04
- PowerTOST and Sample size estimation for R-SABE Dr_Dan 2014-04-18 19:29
- PowerTOST and Sample size estimation for R-SABEAstea 2015-06-25 01:14
- Sample size for scABEL/RSABE vs. conventional ABE d_labes 2015-06-25 09:19
- PowerTOST and Sample size estimation for R-SABE mittyri 2015-06-25 10:43
- PowerTOST and Sample size estimation for R-SABE Astea 2015-06-25 13:54
- Scaled ABE d_labes 2015-06-25 14:04
- PowerTOST and Sample size estimation for R-SABE Astea 2015-06-25 13:54
- PowerTOST and Sample size estimation for R-SABE kumarnaidu 2014-04-18 13:38
- EMA’s ABEL Helmut 2014-02-07 14:21