Clemson University, Fall 2004 
Course Info
Syllabus: (PDF PS) 
Homework
HW #1 Due Th 8/26 PDF PS HW #2 Due T 8/31 PDF PS HW #3 Due F 9/03 PDF PS HW #4 Due F 9/17 PDF PS HW #5 Due Th 9/23 PDF PS HW #6 Due T 09/28 PDF PS HW #7 Due Th 10/21 PDF PS HW #8 Due T 11/16 PDF PS 
Exercises
Exercise #1 (due T 09/07) PDF PS Exercise #2 (due F 09/10) PDF PS Exercise #3 (due F 09/17) PDF PS radioflat.dat radiosteep.dat Exercise #4 (due F 10/15) PDF PS Exercise #5 PDF PS Exercise #6&7 PDF PS 
Readings
For Th 8/26, give a qualitative nondetailed read to the 1990 paper on linear regression in astronomy by Isobe et al (a PDF version of which is below). As usual, the detailed mathematical details are not particularly important. I want you coming away with the general idea behind ordinary least squares, when the approach is valid, and a sense that you probably had no idea it was such a muddy area requiring thought before utilizing. PDF file For T 8/24, read J. Patrick
Harrington's (U MD) basic statistics primer/review. This will either
remind you or introduce you to the basic key distributions in astronomy,
and some of the salient results. Note in particular the use of moment
equations to define the mean and variance. As we've said, these are
used in other areas of physics (esp. statistical physics), and we may someday
see more when we take moments of the radiative transfer equation in a future
stellar atmospheres course. I also want you to be sure that you don't
drift off when reading the section on least squares fitting. First,
note the math and derivations look downright ugly with a quick glance,
but I think even with a tiny bit of attention they are really easy to follow.
Second, note that this is a generalized technique you could apply a "least
squares" approach to fitting any function. Third, note that this
is a socalled "maximum likelihood estimate" of a bestfit line.
Maximum likelihood techniques generally maximize likelihood by minimizing
the variance (the sum of the squares of the residuals between the data
y values and the fitted value at some x) with respect to relevant parameters
of the fitting function (for a line, the slope and zeropoint).
For the first class Th 8/19
Read sections 1 and 2 of this introductory review of Bayesian inference
by Tom Loredo. Your focus should be on a (perhaps initially
nebulous) qualitative understanding of Bayes theorem and how Bayesian inference
differs from the "frequentist" approach. You are encouraged to look
at some of the sample applications in later sections of the article (they
might be useful for homework) particularly the subtle art of establishing
prior probabilities but we will try to establish "ok, so how do I actually
use this?" in class.
