Sample Size and Power: Phase 1 Trial Examples Part 2

So, a few of our examples. This short hint, because I always get asked this question. How many humans do I need when I know nothing? Sometimes the basic element has to do more with stability of the estimating equation than with really solving the problem. So, you may run one of these calculations that tells you, you only need three people per arm.

You will need more than three people per arm. Fifteen people per arm or rats per arm, et cetera, tend to be the minimum number. This is a good rule of thumb for early studies. Why is this? Well, when you have 15 -- 12 to 15 values, you actually get a somewhat stable variance equation. I'm not saying it's right.

It could be that you have weird sampling, and the variance is off because of bias, but the equation itself for variance stabilizes between -- sometimes as low as 10, but usually 12 to 15 numbers in it. If you're using a Bayesian analysis technique, sometimes you need at least 70 people per arm before the analysis

is, in fact, stable. A lot of our studies, a lot of our analyses and regressions float to something called the Z test. However, if you have a smaller sample size below 20 or 30, even sometimes below 100, you need to check t-distribution. If you're doing something like logistic regression, principal components analysis, et cetera, you usually need a minimum of 10 participants per variable.

These are individual participants per variable. And in some cases, you may need 100 per variable in order to get stable estimates. This is for like -- the math will spew out of the computer, it doesn't mean the functions are stable. So, there's this balance of wanting to find significant results. What do I need to if I think these are the differences I'm aiming for?