Course Announcement

550.630: Statistical Theory I

Fall 2014

The fundamentals of mathematical statistics will be covered. Topics include: distribution
theory for statistics of normal samples, exponential statistical models, the sufficiency
principle, least squares estimation, maximum likelihood estimation, uniform minimum
variance unbiased estimation, hypothesis testing, the Neyman-Pearson lemma, likelihood
ratio procedures, the general linear model, the Gauss-Markov theorem, simultaneous
inference, decision theory, Bayes and minimax procedures, chi-square methods,
goodness-of-fit tests, and nonparametric and robust methods.

Prerequisites: Probability (550.420 or 550.620 or some such)

Time: MW 3-4:15
Section 1: Th 3:00

Text:
Bickel and Doksum, Mathematical Statistics, Vol I, 2/e, Updated Printing, 2007 
--- (Be sure to get the 2007 "Updated Printing" version of the textbook!) 

Instructor: Carey E. Priebe
Whitehead 201 / Clark 301D
cep@jhu.edu

-----

There are two natural follow-on courses to 550.630: Statistical Theory I.

550.631: Statistical Theory II
is, as the name suggests, a follow-on statistical theory course to 550.630.
B&D (necessarily) leaves quite a few interesting and important topics
unexplored or underexplored.
 (This aspect of B&D seems to disconcert some students.
 Why mention them at all, if they are not to be properly addressed?
 cep response: Would it be better to not mention them at all?
 Perhaps, if this were a course teaching blind adherence to a statistical script
  ["if A, then do a t-test; if B, then ..."].
 But statistical scripts are not the purpose of 550.630.
 The purpose of 550.630 is to begin to *understand* statistical theory and practice.
 And that includes at least being aware of issues, even if their resolution is not clear ...)
In 550.631 we choose some of these underexplored topics for in-depth study.
Students are encouraged during 550.630 to identify individual sentences or paragraphs
from B&D for nomination as candidates for coverage in 550.631.
["B&D" == Bickel and Doksum, Mathematical Statistics, Vol I, 2/e, Updated Printing, 2007]

550.735: Topics in Statistics: Pattern Recognition
"Statistical pattern recognition" is my version of an explicitly prob/stat-centric
"machine learning" or "data mining" course, following naturally from 550.630.
Both the theory and the practice of classification are stressed.
["DGL" == Devroye, Gyorfi and Lugosi, A Probabilistic Approach to Pattern Recognition, 1996]