Magic tricks when well-performed can serve to amaze, entertain, and inspire wonder. They can also serve an important pedagogical purpose in a probability and statistics classroom. Because most magic effects rely on confounding a spectator's perception of probability, they can be used as interactive demonstrations to illustrate important probabilistic concepts. In this talk, several example magic effects are presented that can help to enhance students' understanding of probabilistic reasoning, illuminate subtle concepts in probability theory, and provide the starting point for more formal discussions of probability calculus.
Experiments in which treatment assignments are randomized, such as FDA clinical trials, are considered the gold standard for estimating the causal effect of some intervention (a new drug, a reading program for second graders, a smoking cessation program). However, randomization is often not feasible due to ethical or practical reasons. In those cases the challenge is to do as well as possible with observational (non-randomized) data. Matching methods can replicate two key features of randomization: finding treated (exposed) and control (unexposed) groups who are as similar as possible on observed covariates, and setting up the design without using the outcome. This work looks at the consequences of matching using multiple control groups by establishing a theoretical framework and generating practical guidance from that framework. The method is applied to the evaluation of a school dropout prevention program, where the original treated and control students were significantly different from each other. We look at the use of another source of control students to supplement the original control group and propose a method to adjust for unobserved differences between the two sources of control students.
Health services is an important area of research. Differential access to and utilization of medical care by various groups has yielded important insights into the United States health care system. Dr. Ash has used recent Medicare data to explore the sources of the surprising finding that Blacks, Hispanics and other minorities use more Medicare dollars in their last 6 months of life than Whites.
Researchers interested in measuring people's underlying attitudes (for example, towards their own nation) often collect Likert attitude data. Likert data consist of people's responses--selected from ordered categories expressing different levels of disagreement/agreement--to statements such as `I am ashamed of my nation.' However, people's responses reflect not only their attitudes, but also their response style, which is defined as a consistent and content-independent pattern of response category selection, such as a tendency to agree with all statements. Ignoring response style differences in the analysis can result in biased conclusions about people's attitudes. In this talk, I present a new model for Likert data that allows both attitudes and response style to affect people's responses. The model can be used to study the effects of demographic variables on underlying attitudes, with adjustment for demographic differences in response style. For illustrative purposes, data from the 1995 National Identity Survey (NIS) is used to compare American and British attitudes towards their respective nations.
Using data analysis, we explore three menaces in the United States that seem greater now than a few years ago: The Electoral College, the minority-student achievement gap, and aviation terrorism. In both the 2000 and 2004 Presidential elections, the eccentricities of the Electoral College created widespread disillusionment among Americans. The minority-student achievement gap--more visible than ever since passage of the "No Child Left Behind" law--does not lend itself to simple explanations. And there is scant reason to believe that the terrorists who so stunningly succeeded on 9/11 could not cause further aviation calamities. We discuss these problems, and suggest some steps to deal with them.
GEE and GLMM: Oh my!
Dr. Ken Kleinman,
Associate Professor of Biostatistics, Harvard Medical School.
May 26, 2006, 11:00am, Burton 302
What Scientists and Statisticians Need to Know About the Protection of Human Subjects
Dr. Nicholas Horton,
Assistant Professor of Mathematics and Statistics, Smith College.
May 30, 2006, 9:30am, Burton 302
Thanks to the Smith College Lecture committee, Department of Mathematics and Statistics, Office of the Dean of the Faculty and the Alumnae Association for support of the series.
Applied Statistics Lecture series (2004-2005)
Other 5 college seminars of interest:
University of Massachusetts Statistics and
Probability Seminar Series
Organized by Nicholas Horton.
Last updated May 27, 2006