search

config-i1 / mes

Mixed Exponential Smoothing
0 stars 0 forks source link

Advanced loss functions to implement #33

Open config-i1 opened 4 years ago

config-i1 commented 4 years ago

The analytical counterparts of MSEh, TMSE, GTMSE and GPL are still not implemented. This can be useful for high frequency data / small samples.

config-i1 commented 4 years ago

Also make sure that LASSO and RIDGE work in case of !etsModel and in case of arimaModel

config-i1 commented 4 years ago
  1. LASSO and RIDGE work (whether this is correct, is a different question).
  2. Analytical counterparts of multistep estimators rely on covaranal() function, and can be implemented, when we have proper parametric intervals (issue #39).
config-i1 commented 4 years ago

Analytical multistep estimators are currently not implemented. They need to be done correctly for both Etype="A" and Etype="M", but the covariances for the latter are difficult (although doable). The main issue is nobody cares.