Energy Market Forecasting
This was the first time this class was offered. I covered the topics you’d expect:
- Basic information about energy markets
- Energy market data examples
- ARMA, VAR, GARCH, nonlinear, and machine learning models
I’m absolutely going to make major changes the next time I teach the class. One mistake I made was writing on the electronic screen and then posting all the notes after every class. That led to some students not taking notes, and judging from the projects, some of them had no understanding of the material. To be sure, the better students understand the material really well, but those at the bottom end of the distribution did far worse than students in similar grad courses have done in previous semesters. I will not make the same mistake the next time I teach this class.
Other things I did:
- I used Posit Cloud. It has a lot of promise, but it’s a long way from what I’d consider a good teaching tool. The biggest issue was availability. It’s slow to start, constantly suspends sessions and requires a lengthy restart period, and sometimes we weren’t able to access the site at all. Then there’s the way homework is handled. It would clutter my personal workspace with a homework for each student. I didn’t find the documentation to be that great. It was only at the midpoint of the semester that I learned students were installing packages from scratch on every homework. I solved that by using a template, but it would have been nice if that was clear from the start. Another PITA is that students have to install packages themselves. I don’t understand why I as the instructor can’t install a package during the lecture and have the students be able to access it.
- I put a lot of emphasis on forecast evaluation, forecast comparison, and model selection. I emphasized the importance of accounting for uncertainty about forecasts. I didn’t use formulas that require restrictive assumptions. I did things like computing an out-of-sample MAE and using that to construct an interval. There’s also a lot of coverage of finance-related topics, such as VaR and backtesting.