Regenerate power curves with increasing sample size from 10 to 100 on the following simulations, outer product-ed with the following algorithms.
(Independent AR(1), Correlated AR(1), Nonlinear (our simple version)) X (DCorrX, MGCX, LjungX, wildHSIC, shiftHSIC)
Generate power curves with extinction rate increasing from epsilon to 1 - epsilon for the Nonlinear Dependence process for the following algorithms: (DCorrX, MGCX, LjungX, wildHSIC, shiftHSIC)
Generate power curves with AR component (phi) increasing for the Independent AR(1) process for the following algorithms: (DCorrX, MGCX, LjungX, wildHSIC, shiftHSIC)
Report any issues/enhancements in mgcpy code, and feel free to change figure making code in any way you see fit.
Acquaint yourself with the demo notebook here that generates power curves. https://github.com/neurodata/mgcpy/tree/time-series/demos/time_series
Acquaint yourself with MATLAB code for generating the power curves (data) using wild-HSIC and shift-HSIC, as well as one more skim of the papers to understand what's going on. https://github.com/rmehta004/HSIC https://github.com/rmehta004/wildBootstrap
Regenerate power curves with increasing sample size from 10 to 100 on the following simulations, outer product-ed with the following algorithms. (Independent AR(1), Correlated AR(1), Nonlinear (our simple version)) X (DCorrX, MGCX, LjungX, wildHSIC, shiftHSIC)
Generate power curves with extinction rate increasing from epsilon to 1 - epsilon for the Nonlinear Dependence process for the following algorithms: (DCorrX, MGCX, LjungX, wildHSIC, shiftHSIC)
Generate power curves with AR component (phi) increasing for the Independent AR(1) process for the following algorithms: (DCorrX, MGCX, LjungX, wildHSIC, shiftHSIC)
Report any issues/enhancements in mgcpy code, and feel free to change figure making code in any way you see fit.