Timeline for Is there a reason we need to make a logistic regression linear using the logit?
Current License: CC BY-SA 4.0
5 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Nov 3, 2021 at 19:44 | comment | added | Single Malt | In an answer in the following link, the Stata logistic regression includes “Iteration 6: log likelihood = -17.161893”. Is this information useful in any way doing model building or perhaps even inference? stats.stackexchange.com/questions/144960/… // For logistic regression is it usually convex or not? I thought (possibly wrongly!) that apart from with severe multicollinearity it usually always converges, and so I assumed logistic regression is a convex problem. | |
| Nov 3, 2021 at 12:26 | comment | added | Sextus Empiricus | @SingleMalt my plan is to redo that last graph (which I took from another question) for the case of logistic regression. Maybe that would clarify it better. | |
| Nov 3, 2021 at 12:22 | comment | added | Single Malt | It is useful to know this as this is the “engine” of the method. Is there practical use for the iterations needed for convergence? So for example the Stata glm command with logistic regression option shows for each iteration the log likelihood. Presumably this information is useful, perhaps when comparing two different models faster convergence (fewer iterations) or higher log-likelihood is better. Rephrased, how could this output be used? | |
| Nov 3, 2021 at 10:39 | comment | added | Haitao Du | +1 always enjoy reading your answer! | |
| Nov 3, 2021 at 10:33 | history | answered | Sextus Empiricus | CC BY-SA 4.0 |