SFASA Meeting
Activities |
Presentations
by Student Travel Award winners and Chapter Officer Elections |
Date |
Thursday,
June 20th |
Time |
4:00pm
- 6:00pm |
Location |
UC
Berkeley, Haviland Hall, Room 2 |
Directions
and Parking |
The
Student Travel Award winners are:
Michael
Higgins |
Improving
Experiments by Optimal Blocking: Minimizing the Maximum Within-block Distance |
|
|
Miles
Lopes |
The
convergence rate of majority vote under exchangeability |
|
|
Stephan
Ritter, joint work with Alan Hubbard |
relaxnet and widenet:
Entending the glmnet R
Package with Relaxation, Basis Expansions and Aggressive Cross-Validation |
Speaker: Michael Higgins
Title: Improving
Experiments by Optimal Blocking: Minimizing the Maximum Within-block Distance
Abstract: We develop a new
method for blocking in randomized experiments that works for an arbitrary
number of treatments. We analyze the following problem: given a threshold
for the minimum number of units to be contained in a block, and given a
distance measure between any two units in the finite population, block the
units so that the maximum distance between any two units within a block is
minimized. This blocking criterion can minimize covariate imbalance, which is a
common goal in experimental design. Finding an optimal blocking is an
NP-hard problem. However, using ideas from graph theory, we provide the first
polynomial time approximately optimal blocking algorithm for when there are
more than two treatment categories. In the case of just two such categories,
our approach is more efficient than existing methods. We derive the variances
of estimators for sample average treatment effects under the Neyman-Rubin potential outcomes model for arbitrary
blocking assignments and an arbitrary number of treatments.
Speaker: Miles Lopes
Title: The convergence
rate of majority vote under exchangeability
Abstract: Majority vote plays
a fundamental role in many applications of statistics, such as ensemble
classifiers, crowdsourcing, and elections. When using majority vote as a
prediction rule, it is of basic interest to ask
"How many votes are needed to obtain a reliable prediction?" In the
context of binary classification with Random Forests or Bagging, we give a
precise answer: If err_t denotes the test error
achieved by the majority vote of t > 1 classifiers, and err* denotes its
nominal limiting value, then under basic regularity conditions, err_t = err* + c/t + o(1/t), where
c is a constant given by a simple formula. More generally, we show that if
V_1,V_2,... is an exchangeable Bernoulli sequence with
mixture distribution F, and the majority vote is written as M_t=median(V_1,...,V_t), then 1-\E[M_t] = F(1/2)+
(F"(1/2)/8)(1/t)+o(1/t) when F is sufficiently smooth.
Speaker: Stephan Ritter,
joint work with Alan Hubbard
Title: relaxnet
and widenet: Entending the glmnet R Package with Relaxation, Basis Expansions and
Aggressive Cross-Validation
Abstract: I will describe two
R packages that IÕm working on. relaxnet
applies the idea of the Relaxed Lasso (Meinshausen,
2007) to glmnet models (as provided by the glmnet R package, Friedman et al, 2010), leading to
increased prediction accuracy in certain cases, and fewer false positives (i.e.
fewer noise variables in the final model). widenet adds the capability of applying polynomial
basis expansion to the input data and then selecting a subset of the basis
functions using relaxnet. The intent with both of
these packages is for the user to make aggressive use of cross-validation to
select tuning parameters, and this is encouraged by providing
options to easily parallelize the execution over cross-validation folds.
A preliminary version of relaxnet is available on
CRAN:
http://cran.r-project.org/web/packages/relaxnet/index.html
Slate of SFASA officers for 2013 – 2014:
President: Clinton Brownley President-Elect:
Ruixiao
Lu Past
President: Kit Lau VP BioStatistics: Jing Huang VP
General Application: Megan Price |
Treasurer:
Doris Shu Secretary:
Jacqueline Shaffer Chair
of Short Course: Lu Tian Chapter
Representative: John Kornak Webmaster:
Dean Fearn |