Holistic Doctors, and medicine

As molbiogirl and nator have asked, the way you determine sample size is dependent upon the distribution you are expecting. And if you have any information regarding how prevalent the trait you are testing for actually exists in the population, that will have an effect, too. Any decent statistics book should have a section on estimation and processes to help you determine what your sample size must be.For example, if we have a binomial distribution with a series of n independent Bernoulli trials, we're trying to find the smallest n such that:P(|Y/n - p| < d) = 1 - aIf you want, I can go through all the math involved, but it comes down to, assuming we don't know p:n = z²_a/2p(1-p)/d²But if we don't know p, then we notice that p(1-p) is always less than or equal to 1/4 so in a completely blind situation:n >= z²_a/2/4d²So if you want a sample such that we'll be within 0.05 of the true proportion with 95% confidence, (z_a/2 = 1.96), we get:n >= (1.96)²/4(0.05)² = 384.12Or 385. But if we know that p = 0.25, we get:n = (1.96)²(0.25)(0.75)/(0.05)² = 288.12Or 289.

---

Rrhain

---

Thank you for your submission to Science. Your paper was reviewed by a jury of seventh graders so that they could look for balance and to allow them to make up their own minds. We are sorry to say that they found your paper "bogus," specifically describing the section on the laboratory work "boring." We regret that we will be unable to publish your work at this time.

molbiogirl responds to me:

quote:

---

In order to test whether a "holistic" practitioner could sense an HEF (human energy field) thru TT (therapeutic touch), one would need a sample of either 385 for 0.05 of the true proportion with 95% confidence or 289 for 0.25 of the true proportion with 95% confidence?

---

...and this is where my statistics breaks down. My degree is Applied Mathematics, not Statistics. I don't rightly know enough about the situation at hand to be able to say if we have the appropriate model involved. While it certainly seems to be appropriate that this is modeled by n independent Bernoulli trials, I don't want to insist upon it. For example, this model assumes that all people who can detect this life energy have the same ability to sense it. I'm not willing to make that assumption.But just to clarify, it's 385 people if we don't know what the chance is for any given person to be able to detect it (or if we know that exactly 1/2 the population can). If, however, we know that it's 25% (or, complementarily, 75%), then it's down to 289.If we know it's 10% (or, complementarily, 90%), the sample size drops to 139.If 12 is going to be a valid sample, then p must be about 3% (or, conversely, 97%):12 = (1.96²)[p(1-p)]/(0.05²)
12(0.05²)/(1.96²) = p - p²
p² - p + 0.3 = 0Simple quadratic, solving for p:p = .031 or .969So to the higher question as to whether or not 12 is a valid sample, the answer is: It depends. If we expect something to happen practically every single time (or, conversely, practically never), then we don't need very many trials to have a high confidence. If these practitioners can be expected to be able to detect it with near-perfect ability, then yes, a sample size of 12 is sufficient.

---

Rrhain

---

Thank you for your submission to Science. Your paper was reviewed by a jury of seventh graders so that they could look for balance and to allow them to make up their own minds. We are sorry to say that they found your paper "bogus," specifically describing the section on the laboratory work "boring." We regret that we will be unable to publish your work at this time.