View Single Post
01-01-2013, 02:40 PM
Registered User
DAChampion's Avatar
Join Date: May 2011
Location: Baltimore, Maryland
Country: United States
Posts: 15,775
vCash: 500
Originally Posted by Mathradio View Post
Undergraduate education does matter since that's where graduate schools feed from. A reputable university should be able to do both at a decent level.

I got CS covered into a numerical physics class; we learned stuff about things like chaos as well as numerical methods but applying that stuff was done in computer labs where we were expected to code and yield results from it.

Three of the eight graded labs pertained to astronomy:

- Lab #4 (actually the second graded lab; the first two were non-graded) pertained to planetary orbits in a single-star system with one planet and, later two planets initially placed very close together
- Lab #8 was about stellar radiance (and where we used Fourier series)
- The final project, which was an extension of lab #4, where you have to simulate a stellar system with not one but two and, ultimately, three planets, with a caveat on the planet at 1 AU, which has to be a super-Jupiter whose mass is 2% that of the Sun.

But is the statistics elective from CEGEP (a course that a CEGEP student can choose from a list that contains MV calculus and statistics) useful to do astronomy?
I'm not sure what stats course you did, but presumably it covered the central limit theorem, means, standard deviations, linear least-squares-fitting, poisson distribution, chi-square test, et cetera so it should be useful.

Do you know what a markov chain monte carlo is? Probably not, it's very important and most undegrads don't learn it. However, if you've taken a good stats class you'll be able to pick up a book, understand, and implement the algorithm within a few days or weeks, which is perfectly adequate.

Do you know how to do multivariable least-squares? If not, with a good stats class you can figure it out within a few hours.

A good stats course sets you up to learn and to be able to learn what you need to figure out, and also to have good instincts. Do you get the idea of correlations? Correlations are very important, since most errors are correlated, i.e. if you overestimate a star's temperature, you'll probably be overestimating the metallicity as well.


What you need to know about computers is how to code, how to implement algorithms, and how to use other people's black-box software. Those labs sound pretty good actually.


Aside from all this you also need to know how to write. This was never a block for me, but it's a block for a lot of people I know, the block in fact that prevents them from succeeding. They do a lot of good work then when they need to write it up they might put together a few bad paragraphs in a day.

I presume this won't be an issue for you; you should consider this a major advantage.

DAChampion is online now