(Total Views: 704)
Posted On: 12/12/2020 2:24:31 PM
Post# of 155333
Re: Borel Fields #67992
Borel Fields,
For the p-values I am using the binomial distribution between two independent samples.
The ad-hoc power is calculated with the same numbers as the p-value as if the test had been completed. So, for example, for p=0.04384 I used 23.5% and 37.8% (deaths percentage) respectively. This defines the difference between the proportions.
The formula used in the calculations is the one depicted in the B. Rosner book "Fundamentals of Biostatistics" Brooks/Cole, 2011.
I understand that the majority use confidence intervals rather than post-hoc power calculations to evaluate an adequate sample size but imo this gives us more information because the trial designers had in mind this as a parameter of the design (likely 80) and any deviation will be looked at negatively by the FDA.
For the p-values I am using the binomial distribution between two independent samples.
The ad-hoc power is calculated with the same numbers as the p-value as if the test had been completed. So, for example, for p=0.04384 I used 23.5% and 37.8% (deaths percentage) respectively. This defines the difference between the proportions.
The formula used in the calculations is the one depicted in the B. Rosner book "Fundamentals of Biostatistics" Brooks/Cole, 2011.
I understand that the majority use confidence intervals rather than post-hoc power calculations to evaluate an adequate sample size but imo this gives us more information because the trial designers had in mind this as a parameter of the design (likely 80) and any deviation will be looked at negatively by the FDA.
(0)
(0)
