Multi-level models for nested variables in time dimension
The math of mixed models doesn't care whether the original dimension was space, time or something else entirely. In both time and space, you can have autocorrelation that messes with model assumptions, but it's that autocorrelation that matters more than the physical interpretation of the dimension. All that said, it's incumbent on the user to know what the inferential interpretation of the resulting model is. Methods designed to deal with serial autocorrelation may have a more obvious interpretation. But the question of which model gives you the inferences you need is one that requires the knowledge of your data and research question that only you have. Hope that helps, Phillip
On 11/19/21 18:21, Vitor Vieira Vasconcelos wrote:
Good night, friends!
I have been seeing many multi-level models, using the mixed models'
framework, that use groups nested in "space", such as students nested in
classes, which are nested in schools, and with independent variables for
each of these spatial resolutions.
Then I was thinking whether we could use this same framework to model
variables nested in the time dimension. For example, if we have some
variables sampled at daily resolution, other variables at monthly
resolution and others at year resolution, and we would like to use all them
in the same model to predict a dependent variable at daily resolution.
Basically, I am just thinking about transposing the same framework from
"space" dimension to "time" dimension, and not thinking yet about
autocorrelation or other time-series analyses.
Do you think that these ideas make sense to you?
Best regards,
Vitor Vieira Vasconcelos
+55-31-99331-1593
[[alternative HTML version deleted]]
_______________________________________________ R-sig-mixed-models at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models