@MatthewDrury @GeneralAbrial you two

Unrelated note: unevenly-spaced time series. We have a lot of questions on it here, and on StackOverflow, and it seems the consensus is that "nobody really has a good way to handle them."

It seems the only recent, high profile research on it was by Andreas Eckner, but his whole website seems to have not been updated in a while. And I don't see any other research making use of what he did

Is there some reason I'm missing as to why this isn't an interesting research topic to the stats community?

Or are people just adapting spatial correlation models to 1-D time series applications and calling it a day?

@ssdecontrol I'm wondering whether the problem is addressed somewhere and we're all just blissfully unaware of it. Or it's known to be intractable without imposition of many assumptions

or.... it's a question that's so underdefined that it's a statistical Rorschach test.

last point is interesting

There probably isn't a standard way to adapt/generalize, say, ARIMA to that context

Gaussian processes seem to be one standard-ish way to handle

and dynamic linear models

@ssdecontrol Over here in social science and biostatistics, we don't call unevely spaced time series "time series"; we call them longitudinal data, and employ methods such as mixed models.

Well, there's this jstor.org/stable/2337095

And stat.columbia.edu/~rdavis/papers/pap2.pdf and thankfully an r package cran.r-project.org/web/packages/cts/vignettes/kf.pdf

@ssdecontrol thanks for your answers for many of my questions. I like your insights very much.

Hi everyone :

Can anyone tell me the formal approaches taken to tune in the parameters for a RandomForest

*regressor !

@wolframalpha That's a fine question for the main site if you can't find what you need with a search.

But, it's unclear what you mean by "formal"

*formal == Thumb rule ?

I would consider those very different.

I was looking through this very useful list of model formulas in R conjugateprior.org/2013/01/formulae-in-r-anova and it gives an example of "A is random, but B is fixed and B is nested within A", with formula Y ~ B + Error(A/B). I am confused by this example.