Skip to content

Latest commit

 

History

History
22 lines (11 loc) · 1.15 KB

post-workshop-thoughts.md

File metadata and controls

22 lines (11 loc) · 1.15 KB

Post-workshop thoughts. NYC 2018

Day 1. Putting tscv with ETS at end of day a good idea, but give them time to do it. Also explain about writing functions, etc.

Day 2. Great day. Doing STL after ETS works much better. Perhaps give them time to explore STL on its own without forecasting first.

Day 3. Dynamic regression was a hit, but I need a simpler first exercise. The motel exercise is too difficult (due to quarterly CPI problem). Perhaps get them to do an extended uschange, or something else?

Perhaps drop the whole section on needing non-stationary data. The software works ok anyway. Also drop the estimation stuff. Just concentrate on the models and their interpretation.

Don't forget to do the wrap up!

Spending 30-60 minutes on questions, or just going through the extras slides worked well. But add in ensemble combination benchmark post, and something on M4 competition results.

Maybe drop seasonal dummies. Not useful. Also drop linear trend.

Reorganize so I do the Fourier example after introducing Fourier terms.

Drop transfer function slides, just show Quotes/Advert as an example

Give them time to use stlf on multiple seasonal time series.