By the end of yesterday, I had 5 transit light curves, one for each planet in the system I'm studying, Kepler-186. By the end of today, I had a list of all of "Good" points and a list of all of the "Outliers". An outlier is a point that is so far away from the average of the other points that we can tell it isn't actually a reflection of what's happening in our target. It might be the result of an instrumental error, or a strange spike in stellar luminosity, or a number of different things. Whatever it is, it isn't part of the transit data I'm interested in, so I don't want it. My other advisor, Dr. John Johnson, explained it like this:
Getting these lists of good and "bad" points was a multi-step process. First, I had to identify all of the points that occurred outside of at least one transit for any of the planets. In other words, if a point existed outside any transit for planet b, planet c, planet, d, planet e, or planet f, I made a note of it. It is from this "Out of Transit" list that I can begin to remove outliers.
Why did I have to first remove the in-transit points? Well, the point of a transit is that its points exist par from the mean amount of light coming from the star. If I removed outliers from the entire dataset, I would also remove most of the transit points, but that's the bit that actually leads to scientific results!
Removing outliers is more-or-less straightforward. Using the "Out of Transit" data points, I find an average flux value.
For our data, this value is essentially 1. I then find an acceptable sigma value using a "mad" function from the interwebz. The function finds the Median Absolute Deviation, and calls itself a more robust form of finding the standard deviation. This is necessary because a standard deviation function is too influenced by the outliers. (Those things really do cause so much trouble!)
The last step is to flag, or take note of, all the points that are a certain number of sigmas away from the mean. I set that certain number equal to 3, because anything less than that means removing way too much of the data. (See this wikipedia page to better understand why.)
But after all of that, I'm left with this plot!
Please excuse the awful y-axis labels. I'm figuring out how to get rid of that "+9.96e-1" bit.
You can even see some of the transit points in this plot! They're the black points that fall way outside the mean.
That was one part of my day. After that, I spent another two hours researching galaxy mergers. Well, I spent two hours doing things that will eventually lead to learning more about galaxy mergers. I essentially spent that time reducing Herschel data and playing around with the output fits files in python. I say "playing around" because I haven't actually managed to do anything too useful yet. But I will soon! And when I do, I'll write about it here.
No comments:
Post a Comment