search

gragusa / CovarianceMatrices.jl

Heteroskedasticity and Autocorrelation Consistent Covariance Matrix Estimation for Julia.
Other
53 stars 9 forks source link

Update kernel for QuadraticSpectralKernel #28

Closed lbittarello closed 6 years ago

gragusa commented 6 years ago

I think we have now another problem with the DataFrames 0.11.x which does not define DataFrameRegressionModel. Test pass for me locally with DataFrames 0.10.x.

This means that I have to go through the refactoring before we can merge this. I will open an issue soon with my thought. It would be helpful if @lbittarello and @Nosferican weigh in ....

Nosferican commented 6 years ago

StatsModels provides the struct so, using StatsModels: DataFrameRegressionModel Fixes it while it changes the DataFrames dependency to StatsModels. The issue is GLM hasn’t updated yet to the DataFrames 0.11 and StatsModels 0.2 regime. I played around with shifting the DataFramesRegressionModel to inherit a level higher to StatsBase.RegressionModel and it works after defining a few methods (for it, didn’t test it directly with GLM though). I developed the VarianceCovarianceEstimators under MIT which uses a nice coding framework you are free to look at to adapt the package.

StatsModels will eventually move out of a wrapper to inheritance so it might be a good time to make the change. GLM will most likely make the change at this iteration. Work it with upper bounds in the versions.