Timeline for Logistic regression to adjust for confounders in treatment effect estimation: when is my model satisfactory?
Current License: CC BY-SA 3.0
5 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| May 29, 2017 at 13:48 | vote | accept | Konrad | ||
| May 16, 2017 at 7:59 | comment | added | Konrad | I am unaware of the way "spline methods" work. I've read a lot about the term on StackExchange but I'm not sure how to implement that in a propensity score stratification or matching method. | |
| May 16, 2017 at 7:57 | comment | added | Konrad | I'll also be taking a look at your references. By the way, I have very large volumes of data, if that helps, though I'm trying to measure pretty small treatment effects... | |
| May 16, 2017 at 7:57 | comment | added | Konrad | Thank you Bjorn. I have read a lot about propensity score matching and stratification. I have tried the latter by running a regression with treatment as the outcome and some confounding variables as predictors. I did not find very satisfactory results with this method either, however. Most of my variables are dichotomous and I think that had a pretty negative effect on my stratas (propensity stores quintiles), which in the end did not seem sufficiently balanced. Could this be because my propensity score prediction was not sufficiently robust? | |
| May 16, 2017 at 7:41 | history | answered | Björn | CC BY-SA 3.0 |