home

Research Methods Seminar

9/18/09 - Dr. Michael Dougherty, PSYC/NACS, The lack of robustness of robust statistics to transformation and
a potential solution

A common problem in the behavioral and social sciences is the failure of data to conform to the assumptions of standard least-squares (LS) estimation techniques. Three common approaches to dealing with violations of assumptions include (a) ignoring the assumptions and hoping the statistic is robust to the violations, (b) transforming the data to make it look like it meets the assumptions, and (c) utilizing so-called robust estimation methods. In this talk, I describe a fourth approach to dealing with messy data based on a new statistical modeling approach – the General Monotonic Model (GeMM). GeMM dispenses with least-squares estimation techniques and instead utilizes information contained in the rank orders. Using real data drawn from 3 different domains of psychology (implicit attitudes, cognitive ability, and risky decision making) I illustrate (a) that GeMM provides more stable results than LS techniques, (b) GeMM yields solutions that are relatively insensitive to transformation whereas LS procedures do not, and (c) GeMM has superior statistical power than LS procedures when assumptions are violated.

********************************************************************

9/25/09 - Stephen David, UMD: Institute for Systems Research, *Spectro-temporal representation of sound in human electrocorticography (ECoG)*

10/2/09 - Dr. Joseph Dien, UMD-CASL

10/09/09 - Dr. John Polich, Scripps Inst. in CA, Everything You’ve Always Wanted to Know
about P3a and P3b But Were Afraid to Ask

10/23/09 - Meghan Graham, NACS

10/30/09 - Hui Wang, NACS, Feasibility of Inhibiting Beta site amyloid precursor protein
cleavaging enzyme 1 (BACE1) for the Treatment of Alzheimer’s disease

11/6/09 - Alexia Nunez, NACS, Adult neurogenesis and behavior in the olfactory bulb

11/13/09 - Sharona Atkins, NACS

11/20/09- Leslie Hainley, NACS

12/4/09 - Erika Hussey, NACS