(Total Views: 691)
Posted On: 10/06/2020 6:11:28 PM
Post# of 154120
Dvb,
In cases where there are two groups each with a yes or not outcome (as in success or failure, or live and death) it is commonly to use the binomial distribution as opposed to a continuous distribution.
The good news is that, if we use this test between two binomials (one for LL and the other for placebo), the p-values are significantly lower than the T-Test with unequal variance
.
Without starting a rabbit-hole discussion on mathematics and statistical methods, suffice to say that the p-value for 22-23 deaths LL/Placebo is very good for trial stoppage (I get 0.00482 for a 2-tailed test).
So we are all good with 22/23 Please note that this is a very high bar (16.9% deaths in LL and 35.4% deaths in SOC).
Sleeping uneasy nowadays and checking my telephone every 5 minutes from 5:30 A.M. waiting for the 6:00 A.M PR announcing significant data results obtained
Severe case of COD and a little of greed, I must admit.
In cases where there are two groups each with a yes or not outcome (as in success or failure, or live and death) it is commonly to use the binomial distribution as opposed to a continuous distribution.
The good news is that, if we use this test between two binomials (one for LL and the other for placebo), the p-values are significantly lower than the T-Test with unequal variance
. Without starting a rabbit-hole discussion on mathematics and statistical methods, suffice to say that the p-value for 22-23 deaths LL/Placebo is very good for trial stoppage (I get 0.00482 for a 2-tailed test).
So we are all good with 22/23
Please note that this is a very high bar (16.9% deaths in LL and 35.4% deaths in SOC). Sleeping uneasy nowadays and checking my telephone every 5 minutes from 5:30 A.M. waiting for the 6:00 A.M PR announcing significant data results obtained
Severe case of COD and a little of greed, I must admit.
(9)
(0)
