BEproff Senior Russia, 20160607 21:29 Posting: # 16408 Views: 2,420 

Hi All, Let's say I have results of a BEstudy: CI 0.851.01, n=30, CV=22% and I want to repeat it )) Is it possible using PowerTOST to calculate possiblity that upper and lower limits will be higher and lower than in previous study? Thank you in advance ) 
ElMaestro Hero Denmark, 20160607 22:21 @ BEproff Posting: # 16409 Views: 1,933 

Hi BEProff, » Is it possible using PowerTOST to calculate possiblity that upper and lower limits will be higher and lower than in previous study? I don't know if PowerTOST provides the functionality but it isn't particularly difficult on the general level. The power is calculated as the probability difference P(point est. lower than some High limit)  P(point. est. lower than some Low limit), where the P's depend on sample size CV and GMR and actual limits. But in practive you'll easily find yourself in trouble. If you have a CI of 0.85 1.01 from a prior trial then your GMR is likely somewhere in between, but you don't have a solid basis for saying it is the point estimate (=sqrt(0.85*1.01)). So you might not have a good GMRvalue to plug into the calculations of power for the next trial. — I could be wrong, but… Best regards, ElMaestro A potentially biased estimator may be a relevant estimator. The case of REML speaks volumes. 
BEproff Senior Russia, 20160607 23:11 @ ElMaestro Posting: # 16410 Views: 1,933 

Hi ElMaestro, I will try to reformulate the question :) If my prior results show 0.811.02 I will think that the current formulation is risky for the 2nd study but I don't have any figures to assess the risk. Sqrt(0.81x1.02) returns 0.909...does it give me anything useful for brain? Thnx 
ElMaestro Hero Denmark, 20160607 23:52 @ BEproff Posting: # 16411 Views: 1,918 

Hi BEproff, » I will try to reformulate the question :) Sorry if I did not catch your point. Happens very often Being equipped with a walnutsized brain I am what is typically called an anatomical, cerebral and intellectual anomaly. It does pose some social challenges too. I am obviously born 150 years too late. In 1866 I could probably have been a star in a freak show somewhere. » If my prior results show 0.811.02 I will think that the current formulation is risky for the 2nd study but I don't have any figures to assess the risk. » » Sqrt(0.81x1.02) returns 0.909...does it give me anything useful for brain? Then 0.909 is your point estimate. It would be good if the true GMR is 1.02 and it would be bad if the GMR is 0.81. With the usual assumptions blahblah 1.02 is as likely as 0.81. 0.909 is probably your best guess for the time being, right? People would call that the GMR of maximum likelihood. There was a crazy scientist a few years back who published something about two studies in a row. Something with power, type I errors, and how one can make use of the first study info to plan the second trial. There is a link to the paper (free) here. Try and see if the answer to your question might be in the paper. However, I'd like to stress that the most important conclusion that was made and this is hardly news if your think about it is that when there is uncertainty about the GMR then it is not often wise or reasonable to base any planning on its estimate. And that's how a small pilot trial easily is about as good as no study. — I could be wrong, but… Best regards, ElMaestro A potentially biased estimator may be a relevant estimator. The case of REML speaks volumes. 