Sample of Past Meetings Descriptions:

Randomized Experiments in the Social Sciences

In this presentation, Temple University's own Kevin Arceneaux, associate professor of political science, will discuss experimental methods in both laboratory and field settings. Using examples from his own work, he will highlight the benefits, limitations, and challenges involved in using experimental methods to answer social science research questions.

Dr. Arceneaux investigates the effects of campaign advertisements and endorsements, the factors that facilitate and limit the persuasiveness of political rhetoric, and the impact of the hyper-choice media environment on public opinion. Risking oversimplification, one could say he watches people watching political TV. In each of these projects, Dr. Arceneaux aims to apply theories developed in psychology, evolutionary psychology, and neurobiology to the study of political phenomena.

Come to the lunch, and see what else you could worry about if you're weren't worrying about endogenous variables.

Please RSVP to ssdl@temple.edu to let us know if you plan to attend. If you have any questions or suggestions for future Brown Bag topics, email the SSDL or call us at 215-204-5001.

Using PhilaMetroMapper on The Philadelphia Metropolitan Indicators Project Web site.

Researchers studying communities in the Philadelphia Metropolitan Statistical Area now have online access to comprehensive administrative data on a municipality by municipality basis. This presentation demonstrates easy ways to both access the data and create maps and graphs that illustrate findings in an engaging and intuitive manner. This unique tool aggregates data into useful geographic units, providing a new, user-friendly entry point to data that has previously been unavailable or very difficult to access.

Using Multilevel Models for Longitudinal Data Analysis

Social scientists heavily use longitudinal data – multiple observations collected over time from the same subjects – to make causal inferences and to describe human development.  Unfortunately analyzing longitudinal data is not as straightforward as analyzing cross-sectional data.  Multilevel modeling (also known as hierarchical linear modeling) is a flexible data analysis technique well suited for longitudinal designs.  In this brownbag, Josh Klugman (Sociology) introduced attendees to the logic of multilevel modeling and provided an overview of the different longitudinal applications of this technique.

Verbalizations While Reading and Learning 

For several years, Dr.  Jennifer Cromley has worked with data that are verbalizations by participants while learning or while reading. This think-aloud protocol methodology yields counts of coded verbalizations.  This methodology is widely used in cognitive and educational psychology. Typically, participants differ in the total number of coded verbalizations. Therefore, researchers convert the raw frequency data to proportions of utterances (or behaviors, etc.) and analyze these proportions. The data analytical quandary for this type of research is that all of the typically-used methods have very low statistical power. The question is “which statistical technique is better suited for this analysis?

Quantifying Geographical Effects on Human Behavior 

There are a number of researchers at Temple who are trying to model human behavior at an individual level (i.e. not aggregated) based on geographic characteristics of where individuals reside: whether someone commits a crime, makes healthy lifestyle choices, etc.  These analyses are challenging, not just because of the technical aspects of choosing the right analytical technique for spatial data, but because of the mismatch between theories of how environment affects behavior and our ability to quantify environmental characteristics and their relationship to individuals.  Sometimes researchers face negative results – perhaps because of the mismatch between theory and implementation. Jeremy Mennis, Associate Professor at the Department of Urban Studies will invite a panel to discuss ways we can more accurately match substantive theories of the affect of geographic characteristics on human behavior with the available representational and analytical techniques necessary to adequately test such theories.

Longitudinal Data Analysis.

A team of researchers from the Psychology Department who are studying juvenile offenders will present a research problem requiring longitudinal analysis. Other researchers attending the session, including several with extensive experience with longitudinal analysis, will offer suggestions regarding approaches to this problem.

Joanna Lee, Graduate Student with Larry Steinberg, Psychology Department, will be presenting a missing data problem affecting her analysis of the relationship between police maltreatment and psychological distress experienced by African-American youth involved in the juvenile justice system.

Download more detailed description of the session. More background info.

 

 

 

The Center for Statistical and Information Science at Temple University compiled this information on statistical and data analysis resources at Temple University.  This site contains a listing of courses offered across Temple University that cover some aspect of statistics, experimental design, data collection, data analysis or database programming.