Timeline for Sample size for binomial distribution for rare events
Current License: CC BY-SA 4.0
6 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Oct 6, 2020 at 19:04 | vote | accept | user6883405 | ||
| Oct 6, 2020 at 14:41 | comment | added | EdM | @user6883405 elaborated more on why the normal approximation is OK for most practical power calculations. | |
| Oct 6, 2020 at 14:35 | history | edited | EdM | CC BY-SA 4.0 |
elaborated on normal approximation
|
| Oct 5, 2020 at 21:46 | comment | added | EdM | @user6883405 if you look at the (freely available) chapter by van Belle linked in the answer, you will see that the numerator of 4 in that rule of thumb comes from dividing a more generic numerator of 16 (for 2-sample comparisons) in Table 2.1 by a factor of 4. For 90% power, use instead a numerator of 21/4 or 5.25; for 95% power, use 26/4 = 6.5; for 97.5% power, 31/4 = 7.75, based on values in Table 2.1. More on other approaches when I get a chance later. | |
| Oct 5, 2020 at 19:48 | comment | added | user6883405 | Thank you! How might I modify that rule of thumb for a higher level of power than 80%? I found a binomial sample size estimator implemented in R (binomSamSize) and used the Fosgate estimation approach. The actual Fosgate paper (Fosgate 2005) mentions that it is a less conservative approach than the Clopper–Pearson estimation approach. Are you aware of a Clopper–Pearson approach to estimating sample size in R? | |
| Oct 5, 2020 at 19:30 | history | answered | EdM | CC BY-SA 4.0 |