Timeline for Coding the likelihood function for logistic regression
Current License: CC BY-SA 4.0
11 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Sep 21, 2022 at 8:29 | comment | added | idlatva | @Noah thanks for your reply! I wrote some code with this first but wasn't sure if it took one row of the matrix X at a time. But compared them now and saw that the answers were equivalent. So thanks for the insight! | |
| Sep 21, 2022 at 8:27 | comment | added | idlatva | @JohnMadden Thanks for you answer and thanks for the correction, I didn't know that you could check the function so this was really helpful! | |
| Sep 21, 2022 at 8:01 | vote | accept | idlatva | ||
| Sep 21, 2022 at 3:48 | history | edited | Ben | CC BY-SA 4.0 |
Edit to improve clarity, notation and question title and tags
|
| Sep 21, 2022 at 2:57 | answer | added | Ben | timeline score: 3 | |
| Sep 21, 2022 at 2:48 | comment | added | nwaldo | @JohnMadden could you write your suggestion out as a full answer? | |
| Sep 21, 2022 at 0:10 | comment | added | Noah |
FYI, it will be faster (and clearer) to just do p <- 1/(1+exp(-X%*%beta))). You can always benchmark it against R's glm() function.
|
|
| Sep 20, 2022 at 23:55 | comment | added | John Madden | Good question! However "I have no way to check if my function is correct" is not quite true :) The best way to check an implementation of a likelihood function (and the only way to check if you've developed a new likelihood) is to simulate data with known parameters, put the likelihood and the data in an optimizer (like R's "optim" with method="BFGS"), and see if you can recover your true parameters. | |
| Sep 20, 2022 at 23:17 | history | edited | idlatva | CC BY-SA 4.0 |
added 63 characters in body
|
| S Sep 20, 2022 at 23:13 | review | First questions | |||
| Sep 20, 2022 at 23:49 | |||||
| S Sep 20, 2022 at 23:13 | history | asked | idlatva | CC BY-SA 4.0 |