Timeline for How to compute the probability associated with absurdly large Z-scores?
Current License: CC BY-SA 3.0
7 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Aug 3, 2011 at 13:46 | comment | added | whuber♦ | I posted it in a separate reply. | |
| Aug 2, 2011 at 22:15 | comment | added | Iterator | @whuber, can you post the Mathematica code for that? :) I haven't seen Mathematica in 15+ years, and never for this kind of purpose. | |
| Aug 2, 2011 at 22:01 | comment | added | whuber♦ | Incidentally, the Wikipedia reference to that approximation is broken. Mathematica finds, though, that the relative error (1 - approx(x)/erf(x)) behaves like the reciprocal of $2 \exp(x^2+ 3(\pi-4)^2/(8(\pi-3)))$. | |
| Aug 2, 2011 at 21:55 | comment | added | Iterator | Thanks. I didn't notice that there was Javascript-based support for using TeX, which made the difference in writing that out. | |
| Aug 2, 2011 at 21:49 | history | edited | Iterator | CC BY-SA 3.0 |
math formatting
|
| Aug 2, 2011 at 21:33 | comment | added | whuber♦ | Some amplification of this would be welcome, for two reasons. First, it's best when answers can stand alone. Second, that article writes ambiguously about the quality of the approximation "in a neighborhood of infinity": just how accurate is "very accurate"? (You implicitly have a good sense of this, but it's a lot to expect of all interested readers.) The stated value of ".00035" is useless here. | |
| Aug 1, 2011 at 21:46 | history | answered | Iterator | CC BY-SA 3.0 |