- International Business Management
- Information Technology & Marketing
- Entrepreneurship and Innovation
- Tourism, Sports & Recreation
- Financial Services
- Health Care Management
Events
Research Luncheon Workshop Series:
Statistical Methods for
Advancing the Research Agenda
March 30, 2005
11:30 – 2:00
391 Speakman Hall (CSPD)
RSVP by March 23 to Julie Fesenmaier
juliefes@temple.edu
Presenters:
Francis Hsuan, Ph.D.
Department of Statistics
" Bridging Correspondence Analysis and Poisson
Regression"
Both Correspondence Analysis and Poisson Regression are useful
tools to analyze categorical data. The theory (singular value
decomposition, generalized linear model) and practice of each
method have been well documented. In this talk, we illustrate
both approaches using a published set of marketing survey data,
and then compare the two approaches. We will also explore the
possibility of combining the two methods in one framework.
Damaraju Raghavarao, Ph.D.
Chair, Department of Statistics
"Planning of Choice Experiments"
Several factors (or, attributes) contribute in choosing a product
or procedure. If the attributes are related to benefits/costs,
one looks for maximum benefits and minimum costs. However some
trade-offs are needed . Conjoint analysis using fractional factorial
experiments are commonly used for this purpose and in the widely
used choice sets, dominating or dominated profiles may occur.
We are proposing methodology avoiding dominated and dominating
profiles. We will also discuss the exchange rate for the attributes.
Sanat Sarkar, Ph.D.
Department of Statistics
" A New Approach to Controlling Type I Errors
in Simultaneous Testing of a Large Number of Hypotheses"
A new statistical measure, known as the FDR (False Discovery
Rate), providing an overall Type I error rate in simultaneous
testing of a number of hypotheses was introduced in 1995. Statistical
methods controlling this measure are more powerful than the traditional
methods. Although it is now one of the most commonly used statistical
ideas in modern gene-related research, it has the potential to
be useful in any statistical investigation involving large number
of hypotheses.
