Timeline for Algorithm to take samples from binomial data
Current License: CC BY-SA 3.0
5 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jan 16, 2014 at 13:00 | vote | accept | Youloush | ||
| Jan 14, 2014 at 22:11 | comment | added | whuber♦ | To amplify on @Glen_b's point: after drawing a value from a geometric distribution, you randomly make a keep-or-discard decision with a suitable probability determined by the ratio of the geometric mass to the correct probability mass. Because the multiplier $M$ is close to $1$ and in the vast majority of initial draws the geometric approximation is good, there will be few times any value is discarded. Thus, for little more than the cost of sampling from a geometric distribution, you sample exactly from the intended distribution. | |
| Jan 14, 2014 at 22:04 | comment | added | Glen_b | @Youloush What is it you think is inexact here? (Note that the use of approximation before rejection sampling doesn't make the result inexact.) | |
| Jan 14, 2014 at 14:05 | comment | added | Youloush | Thanks for the idea, quite elegant, not exact but interesting nonetheless. | |
| Jan 13, 2014 at 17:47 | history | answered | whuber♦ | CC BY-SA 3.0 |