Timeline for can you calculate the uncertainty on predictions of y(x) if you regressed x(y)?
Current License: CC BY-SA 4.0
4 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jan 28, 2021 at 7:50 | comment | added | user6376297 | I will post a related question, because there is another issue with this, which I have not touched upon here. | |
| Jan 27, 2021 at 18:44 | comment | added | user6376297 | OK, so that's what it's called - thank you! I tried both approaches, and indeed, when I model $y(x)$, the residuals on the back-calculated $x$ are biased, really all over the place, to the point the method is not viable. Today my boss told me that he would like the uncertainty on $x$ to be smaller when $x$ is smaller, as small $x$'s are associated to 'items' we 'care' more about. So I thought: why not model $\frac 1 x$? It sort of does what I wanted, but the error on $y$ is really too large for this to work. So it's quite a puzzle. | |
| Jan 26, 2021 at 14:14 | comment | added | whuber♦ | See our posts on inverse regression. | |
| Jan 26, 2021 at 12:54 | history | asked | user6376297 | CC BY-SA 4.0 |