Project Luther - Presentation and Wrapup
Project Luther Presentations were today!
Project Luther is nearing completion.
Everyone presented this morning. (well… almost everyone… Amy has been out sick the last couple days).
I feel into an odd trap with this project. This was a two-part project. The first parts was web-scraping. This part can be time intensive and indeed rather frustrating as you deal with odd issues. But nonetheless, it’s fairly straightfoward. You set out with a rather specifc task and target: get THAT data. The second part was far more murky. Sure, it may be that linear regression could be viewed as systematic. You do X, Y and Z, possibly iteratively. But this would ignore the primary difficulties.
This part, the linear regression part, is more akin to the “Science” in “Data Science” than “Data”. You are essentially running an experiment. The pattern you seek may simply not exist. You may end up with a null result.
You do need to do a good job ensuring you’ve done your due diligence in attacking the problem. And you need to develop whatever insight you can, even if the main goal of a predictive model isn’t fulfilled.
As I worked on the presentation, I would occasionally need a chart. That might require rerunning a function or model. And a presentation slide may need an explanation of an issue. To better explain something, is even more investigation needed? Can you tweak the model this way or that to learn more or see anything differently? In short, you could end up right back in a never-ended cycle chasing modeling.
From just the point-of-view of the presentation itself, I chose to continue with reveal.js. I did indeed further develop my skills there. Since it essentially IS a website, this lets me learn more about CSS as well. I may eventually add in some custom javascript for my own desired effects. So using reveal.js for the presentations is preparing me for the upcoming focus on things like D3. For example, on several slides I employed “flex” for placement on the slides.
I also made use of the “export to PDF” aspect of reveal (rumor was the instructors wanted a PDF). It does work, after a fashion. But it did something hilarious with the panoramic background. I’m not entirely sure, but it seems it used the vertical offset - as if the PDF pages were constantly going “down” rather than “right”. Tis a minor thing. A more serious thing is that I used the heirarchical organization of reveal to slip in extra, appendix-like, material into the slide presentation. I didn’t actually use these extra hidden slides. Indeed, they were hidden so well I don’t think anyone realized I had the extras. But they aren’t hidden when exporting to PDF. They’re unfolded and put back into the sequence. Interesting.
The feedback I was given on the presentation chided me for focusing too much on WHAT I did and not WHY. That is, I’d lost view of the Storytelling aspect of the presentation and project.