Timeline for Bang and Robins doubly-robust estimator biased and with large variance?
Current License: CC BY-SA 4.0
6 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| May 11, 2023 at 22:58 | vote | accept | Lachlan | ||
| May 11, 2023 at 22:58 | history | edited | Lachlan | CC BY-SA 4.0 |
fixed error
|
| May 11, 2023 at 16:07 | answer | added | Noah | timeline score: 2 | |
| May 11, 2023 at 12:53 | comment | added | Lachlan | Interesting, that's a good observation. I'd tend to think that including an intercept in a statistical model is generally good and accepted practise, even in the case of centered Y. So I'm surprised by that result. Do you have an interpretation of why that intercept parameter might be causing issues? | |
| May 11, 2023 at 8:14 | comment | added | ehudk |
Thanks for the reproducible example. I reran your code, and the dr_mod seems to be overly sensitive to the intercept specification (your data simulation does not include one). If I just add -1 to the formula RHS (lm(Y ~ A + L + dr_weight -1, data = d)), I get mean: 0.000793 and sd: 0.0808.
|
|
| May 11, 2023 at 5:28 | history | asked | Lachlan | CC BY-SA 4.0 |