Lecture on forecasting

[trvrb]

Put together lecture on forecasting. This extrapolates from previous lectures on serology, timeseries and phylogenetics. Include both epi and evolutionary components. We have this down for mostly Sarah, but with some input from me.

https://github.com/trvrb/sismid/tree/master/forecasting

This could get moved to Day 3 if we're running out of time on Day 2. I'd like to give time at the end of Day 2 for groups to prep presentations. The forecasting lecture can be a short (30 min?) lecture.

[sarahCobey]

I'm fine keeping it short. Would be great if we could keep it on Day 2 for cohesion.

[trvrb]

I agree that Day 2 would be better, but I was thinking that after student presentations on Day 3 could also work. Thinking of it as a forward looking lecture.

[sarahCobey]

I drafted a few slides for the forecasting lecture. The material can get dense fast. What do you think?

[trvrb]

Looks good. Though I might aim to be more clear on the standard timeseries approaches to forecasting epidemic dynamics. I think @ntncmch's Ebola work is the best example I can think of this:

http://ntncmch.github.io/ebola/weekly_forecast.html

[sarahCobey]

(Is there a standard? I should look for review papers too.) The Ebola example will make a good warm-up for the Yang model.

On Thursday, July 21, 2016, Trevor Bedford notifications@github.com wrote:

Looks good. Though I might aim to be more clear on the standard timeseries approaches to forecasting epidemic dynamics. I think @ntncmch https://github.com/ntncmch's Ebola work is the best example I can think of this:

http://ntncmch.github.io/ebola/weekly_forecast.html

— You are receiving this because you were assigned. Reply to this email directly, view it on GitHub https://github.com/trvrb/sismid/issues/15#issuecomment-234159587, or mute the thread https://github.com/notifications/unsubscribe-auth/ABI3QnE9W3p6K-Zxib3vX1pCo278wUQYks5qXwHqgaJpZM4JOX-h .

---

Sarah Cobey, PhD Assistant Professor Ecology & Evolution University of Chicago

(617) 756-7204 (cell) sarahcobey (Skype) cobeylab.uchicago.edu

[sarahCobey]

Have you seen a published version of that work anywhere? I only see the arXiv manuscript. The general approach seems similar to work by Yang et al.

[trvrb]

I don't think there's a publication. I think this is the software they'll using:

https://github.com/StateSpaceModels/ssm

which I've glanced through but haven't actually tried to use. I was suggesting this as it's an obviously useful application to public health surveillance.

(Is there a standard? I should look for review papers too.)

I don't think super standard, but it seems common to use either ARIMA type auto regressive models or to do a mechanistic timeseries fit. Not really my expertise.

[trvrb]

I think this is looking pretty good. Going to close the issue. More attention could be paid of course.