This week, apart from the usual teaching and research, I both attended the Department’s annual Lewis Elton lecture, and read a blog post that takes a swipe at Excel. The lecture was given by Prof Adam Reiss (Johns Hopkins) who shared the 2011 Nobel prize in physics with Saul Perlmutter and Brian Schmidt, for showing that the expansion of the universe appears to be accelerating. This surprised a lot of people but appears to be true, not only is the Universe continually getting bigger and bigger but the rate at which it is doing so, is also increasing. This surprise won the Nobel prize.
It was a very good, very clear talk, and well attended by our students. As someone whose job it is to teach them data analysis, it warmed my heart to hear a Nobel prize winner talking a lot about the very careful analysis, mostly of data from the Hubble telescope, that they have done and are doing. I am no astronomer, but I think the basic idea is as follows: In astronomy you can look into the past by looking at distant objects – as it took a very long time for the light you see from them to arrive on Earth or to Hubble, so you can a history of the universe by looking at near objects (as they are in the recent past) and far objects (as they are in the distant objects). And although Prof Reiss skipped over this, I think the speed they are moving (relative to us) is easy due to the red-shift of light from them. So speed at different times can be obtained, which is half the battle.
This leaves the tricky part, which is working out how far a star or galaxy is from us. That is the technical part that Prof Reiss and coworkers clearly are very good at, and work very hard at making as accurate as possible. It is there that they have to demonstrate ninja-like skills in data analysis to quantify both systematic and random uncertainties. His talk was mainly on this. I hope our students left the talk with a renewed appreciation of how important data analysis is.
Which brings me to a nice blog post by Roger Barlow, essentially on why data analysis is better done with code like Python, rather than in Excel. It is a nice post and quite short, so take a look, but I think the executive summary is that if you want to go from simple data to a simple plot in the minimum time, then Excel is good. For anything else, do it with a program, not only is it easier to perform all but the simplest data analysis but if you want to run the same analysis again on another data set or show someone else what you have done, a program is much more useful than an Excel spreadsheet with an embedded plot and some calculations.
I teach with Python Jupyter notebooks that not only have the code for data analysis, but markdown boxes to describe the methodology, which are so much more instructive than some complex Excel sheet full of cells with obscure calculations. And I would be very very surprised if the Nobel prize winning analysis was done with Excel….
Chance & Necessity 