Timeline for R: nlme: can we use varIdent or varFixed to model known variances?
Current License: CC BY-SA 4.0
8 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Nov 9, 2023 at 23:58 | comment | added | NeuroPanda | After some thinking and reading, I think that even using model with just proportions of variance as input, such as with varFix or within classic weighted least square framework might not be such a big problem. Indeed, if there is higher absolute variance in data, this will tend to be reflected in infered variance and in turn in higher standard error of the coefficients, so the estimates will tend to be unbiased ? | |
| Nov 9, 2023 at 20:46 | comment | added | NeuroPanda | Trying the suggested code resulted in error, varWeights requires varFun object as an input | |
| Nov 9, 2023 at 18:47 | comment | added | NeuroPanda | It would be good he can eventually add his comment. Thank you for your help as well. I was thinking maybe some metaregression package could have this covered as well, but I need compatibility with GAMs. | |
| Nov 9, 2023 at 18:41 | comment | added | Robert Long | I agree, the docs are not great to read, but that is my updated understanding! Maybe Ben Bolker will see this, and chime in. He is more of a mixed model expert than I :) | |
| Nov 9, 2023 at 18:32 | comment | added | NeuroPanda | Thanks, but will it treat the values really as absolute variances,not just proportional weights a la weighted regression? Looking into documentation does not help me to make it clear. | |
| Nov 9, 2023 at 17:40 | history | edited | Robert Long | CC BY-SA 4.0 |
added 207 characters in body
|
| Nov 9, 2023 at 17:34 | history | edited | Robert Long | CC BY-SA 4.0 |
added 207 characters in body
|
| Nov 9, 2023 at 17:28 | history | answered | Robert Long | CC BY-SA 4.0 |