I factored out pieces of code that I always found myself using so that it would all live in one spot (here).
This package has
- Static rolling functions
makekernelusing StaticKernels.jl. For small window sizes, this is probably the fastest existing way to do rolling means, sums, etc.
- Variance ratios
kurtfunctions that are faster and more forgiving than
- Autocorrelation function
StatsBase.autocorwas too annoying when you only want the first one, and this is faster. Plus I allow for transformations like
autocorrelate(abs,x)if you want the autocorrelation of absolute
x. you can replicate
Lare your desired lags.
- Conditional correlations
upsidecor(eventually will live in AsymmetricRisk.jl when I give it more love)
loggrowth(x,n)for computing log growth rates over different horizons
perioditeratorfor labelling observations in simulations.
monthtoquarterDate-like helper functions