Week 4

I want to start this week’s blog post with a very important message: I am Autistic. My brain is wired a certain way and that comes with a certain set of problems, mainly with confidence and the need for structure. With Covid-19 wreaking havoc on routine, I was in desperate need of structure in my life and Dr. Davenport has been excellent about building that support for me. In education research, his method is called scaffolding and it’s hugely important for a lot of learners, especially learners like me. He gives clear, concise instructions that are neither too big nor too small that get me to where we both want to go. He doesn’t give me a large project and tell me to wing it, he gives me the necessary steps to get there and that’s the #1 reason why I feel like I’m rocking this DREU project.

At the very beginning of this project, Dr. Davenport asked me what I needed of him as an advisor. I told him that because I’ve never done a project of this scale without working with a partner before, I don’t know the process. (If you can’t tell, I’m big on processes!) Once I complete this DREU, I will have that experience and I’ll be more comfortable. But right now? Please help! I also told him that I need deadlines. I don’t work well with nebulous “when you have time” or “when you get to it” due dates, I need actual dates that I can work toward. It allows me to have that pressure of a deadline as well as the ability to chunk out my time so that I’m working toward a specific goal. Dr. Davenport and I check in near-daily and he always has a next step prepared for when I finish a step, and he gives me weekly deadline goals. This helps me keep moving forward at a pace we are both comfortable with!

Speaking of moving forward, let’s talk about what I learned this week: Box-Least Squares (BLS) Periodogram! Once again a function that attempts to find periods in the light curves, only this time it is using square waves to find patterns in data. Dr. Davenport and I agreed that this is a better model than Lomb-Scargle (LS) and will likely generate the most powerful periods because of how sharp the eclipse dips were. As long as we can get the duration right, we can ge the vertical parts of the square wave to line up with the eclipse–something that can’t happen with a sinusoidal wave! Since the Autocorrelation Function (ACF) also had the added benefit of using the curve itself to find patterns, we decided to save computation time by giving the BLS function the same periods that ACF found to see if the two would agree on the most likely period for the function. Turns out, they did! Well, mostly. Occasionally the ACF or the BLS would pick up the half-period (meaning the actual period is two times the most powerful period) or it doubled the period (meaning the actual period was half the most powerful period).

Now that I had three periodograms it was time to start finding patterns! Only…there were a few hiccups. The BLS function used a different method of determining power, which meant that the values were not between 0 and 1 like the LS and ACF. Most of the light curves that had eclipsing binaries had high BLS values, but not all, and some non-eclipsing binaries also had high BLS values. It was clear that unless I used a lot of magic numbers and if statements, there was no clear way to pick out the eclipsing binaries with just these three functions. (I did create a working classifier, and it worked perfectly! But I doubt it would extend beyond the 60 light curves I was given.) I have to add something else to my classifier and get rid of the magic numbers. Well, get rid of as many as I can anyway! Some magic numbers are simply ‘domain knowledge’ as Dr. Davenport says! (But my computer scientist side still doesn’t like them, so I want to find a more robust way of handling this.)

Up next week: Smoothing the light curve! If we can get rid of the sinusoidal wave action caused by natural variability and/or starspots, we might be able to easily identify significant outliers that tell us there’s an eclipse!

Written on July 10, 2020