Matthieu Gomez

The other thing is that TimedArray does not work for dates in Int, which is common to hold years. This + the inability to have duplicate times makes it sound...

@nalimilan - Stata has a concept of timed DataFrame and timed grouped DataFrame. Once a time variable is defined, doing L.x creates the lagged variable as in this pull request...

@nalimilan that seems way too error prone too me. I think the best option would be to rename the lag function in this package to shift, which is more consistent...

The sorted-lazy approach won’t work for panel data — how would one check that the vector is sorted within each group? Running `issorted` at each `getindex` call would be too...

Is it really cheap to check that `times` is sorted every time `getindex(:x_lagged, i)` is called?

I guess, pragmatically, I am not sure what would be returned by this command: ```julia df = DataFrame(times = [0, 1], x = ["V1", "V2"]) df.y = lag(df.x, df.times) sort!(df,...

@nalimilan The other thing is that, assuming that the underlying times are not mutated seems very error prone to me. Safety seems more important than memory efficiency.

The macro model is just deprecated at this point, not undefined. So I don’t think this is the explanation for the error. Same thing for categorical btw. On Wed, Oct...

What is the issue exactly? On Thu, Oct 15, 2020 at 10:43 AM Jonas Björnerstedt < [email protected]> wrote: > The sum_up function almost works with knitr::kable, when rendering html or...

Fortunately, RecipesBase implemented something recently to work around this kind of issues. Can you write a reproductible example so that I can solve this?