Timeline for Puzzling predicted values in generalized multilevel model
Current License: CC BY-SA 4.0
19 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jul 26, 2018 at 14:02 | vote | accept | Mark White | ||
| S Jul 26, 2018 at 14:02 | history | bounty ended | Mark White | ||
| S Jul 26, 2018 at 14:02 | history | notice removed | Mark White | ||
| Jul 22, 2018 at 23:48 | history | tweeted | twitter.com/StackStats/status/1021180296513957888 | ||
| Jul 22, 2018 at 22:00 | comment | added | usεr11852 |
Ultimately, the random effects part of an MEM is there because we recognise there is some (possibly id-related) structure in our errors/residuals. If we allow as single error parameter of each residual point then the remaining variability is minor. As such our fixed effects are left with next to nothing to model. This results to the very small coefficient you see. (Note that in logistic regression there is no error term so in comparison with the simple LMER the effect will be even more pronounced)
|
|
| Jul 22, 2018 at 21:45 | comment | added | Mark White | @usεr11852 I make that previous comment (forgot to tag you in it) because I tried simulating some data with the same distribution of 1, 2, and 3 responses and still got OK inferences: gist.github.com/markhwhiteii/4987ed886017eadc1785a5cdef298b3d. Granted, these simulated data were generated by exactly how the model is specified, which is undoubtedly supremely optimistic. You can see the distribution of predictions, broken out by whether or not there are people with 1, 2, or 3, 2 or 3, and all 3 observations: s22.postimg.cc/qv80p008h/… | |
| Jul 22, 2018 at 21:31 | comment | added | Mark White | +1 As an answer, could you explain more of the reasoning behind: "The random effect has sucked the life out of your model. Especially for person who answered a single question all their variability is transferred to the random component and treated as noise." What makes this the case? How does only one observation for a person transfer to the random component? And why do you think MCMCglmm works better? | |
| Jul 22, 2018 at 20:33 | comment | added | usεr11852 |
Try using predict( mod_1, random.only = TRUE) and look at estimates using only the random effects, don't the estimate look suspiciously good? Then try using defining mod_0 = glmer(y ~ (1 | id), dat, binomial), don't the estimate look even more suspiciously good? The random effect has sucked the life out of your model. Especially for person who answered a single question all their variability is transferred to the random component and treated as noise.
|
|
| Jul 22, 2018 at 20:33 | answer | added | Michael M | timeline score: 4 | |
| S Jul 22, 2018 at 19:22 | history | bounty started | Mark White | ||
| S Jul 22, 2018 at 19:22 | history | notice added | Mark White | Draw attention | |
| Jul 22, 2018 at 17:33 | history | edited | Mark White | CC BY-SA 4.0 |
deleted 683 characters in body
|
| Jul 21, 2018 at 20:08 | history | edited | Mark White | CC BY-SA 4.0 |
added 186 characters in body
|
| Jul 21, 2018 at 19:59 | history | edited | Mark White | CC BY-SA 4.0 |
added 75 characters in body
|
| Jul 21, 2018 at 19:54 | history | edited | Mark White | CC BY-SA 4.0 |
added 75 characters in body
|
| Jul 21, 2018 at 19:32 | history | edited | Mark White | CC BY-SA 4.0 |
added 1077 characters in body
|
| Jul 21, 2018 at 18:52 | history | edited | Mark White | CC BY-SA 4.0 |
added 30 characters in body
|
| Jul 21, 2018 at 18:36 | history | edited | Mark White | CC BY-SA 4.0 |
added 112 characters in body
|
| Jul 20, 2018 at 19:06 | history | asked | Mark White | CC BY-SA 4.0 |
