Timeline for Predict probability of outcome from continuous variables
Current License: CC BY-SA 4.0
9 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Dec 27, 2019 at 18:54 | comment | added | NorwegianClassic | Thank you! Thought the offset had to be include in some way, but of course it doesn't. I will have a look at your suggestion, it seems to be what I was looking for. | |
| Dec 27, 2019 at 18:50 | vote | accept | NorwegianClassic | ||
| Dec 27, 2019 at 17:41 | comment | added | seanv507 |
offset from what? default_start .. this should be "irrelevant", based on how the model is setup: eg model might be predicted_start = default_start if time < midnight else 3* default_start. if you predict actual start then a prediction interval tells you what is the "plausible" range a new actual_start will be in, taking into account the amount of data you have to estimate the parameters and the typical error size. en.wikipedia.org/wiki/Prediction_interval.
|
|
| Dec 27, 2019 at 16:48 | comment | added | NorwegianClassic | Thanks. As I wrote to Tim: To clear up my reasoning for not using logistic regression, a larger offset does not mean a higher probability of an incorrect prediction and visa versa. That's what makes this problem a bit too tricky for me. | |
| Dec 27, 2019 at 10:30 | comment | added | seanv507 | I think you are asking for prediction intervals. If you use linear regression (eg with nonlinear inputs), you can do this analytically, otherwise bootstrap... | |
| Dec 27, 2019 at 10:10 | answer | added | Tim | timeline score: 1 | |
| Dec 27, 2019 at 8:21 | history | edited | NorwegianClassic | CC BY-SA 4.0 |
added 4 characters in body
|
| Dec 27, 2019 at 8:16 | history | edited | NorwegianClassic | CC BY-SA 4.0 |
added 2 characters in body
|
| Dec 26, 2019 at 20:03 | history | asked | NorwegianClassic | CC BY-SA 4.0 |