Kalman smoother

[mschauer]

The smoother is missing a line P_pred = model.F * P[:, :, i] * model.F' + model.V in the loop.

[mschauer]

Also I think lines https://github.com/JuliaStats/TimeModels.jl/blob/master/src/Kalman.jl#L172 to L175 are wrong and have to be removed.

[milktrader]

Could you do a PR? And how did you discover this? Maybe we can put in a test?

I'll merge it if there is no objection after a week or so, unless someone else wants to merge it sooner

[milktrader]

@ElOceanografo wrote the Kalman code.

[mschauer]

I found it because I implemented it myself and wanted to check my implementation.

By now there are too many implementations, https://github.com/ElOceanografo/StateSpace.jl https://github.com/JuliaStats/TimeModels.jl https://github.com/wkearn/Kalman.jl https://github.com/setzler/JuliaEconomics/blob/master/Tutorials/tutorial_6_noglobals.jl https://github.com/QuantEcon/QuantEcon.jl/blob/master/src/kalman.jl and, well, https://gist.github.com/mschauer/3ffebdb8581f56341a9d

maybe we can do some serious coordination ;-) cc @wkearn, @setzler

[milktrader]

Agreed! I think JuliaStats would be a natural place for this. Whether it stays in TimeModels or gets its own package is a good discussion.

[mschauer]

Let's keep this issue for the mistake and move to https://github.com/JuliaStats/Roadmap.jl/issues/16

[milktrader]

Thanks for doing this!

[GordStephen]

Fixed in https://github.com/JuliaStats/TimeModels.jl/pull/38