Social Science Methodology Symposia

Sponsored by the Department of Human Development and Family Studies, College of Health and Human Sciences

 

April 11, 2014

TITLE:  Thinking Outside the Box: Emerging Methods for Establishing Causality

video opens new window  https://mediaspace.itap.purdue.edu/media/Thinking+Outside+the+BoxA++Emerging+Methods+for+Establishing+Causality/1_0m0zpftz

PRESENTER:  Yaacov Petscher, PhD, Director of Research, Florida Center for Reading Research, Florida State University

ABSTRACT: Statistical simulations, federal funding agencies, and publishers of peer-reviewed manuscripts repeatedly underscore the importance of using multilevel models. When individuals are nested within groups, such models are critical for correct inferences, especially as it pertains to estimating treatment effects in randomized controlled trials. Conventional applications of multilevel models for such experimental designs are typically used for single outcomes or multiple outcomes with a multiple hypothesis correction. Further, such models assume that the average treatment effect is the best estimate for inferring causality. In the present talk, I will focus on emerging applications of multilevel models in education and the social sciences including multilevel structural equation models, linear quantile mixed models, and multilevel models for partially nested randomized controlled trials. Data from prior randomized studies will be presented with an illustration of how to estimate each type of model using software such a SAS, Mplus, and R.

 

TITLE: Opening Up the Blackbox: Design and Analysis of Multilevel Mediation

video opens new window  https://mediaspace.itap.purdue.edu/media/Opening+Up+the+BlackboxA+Design+and+Analysis+of+Multilevel+Mediation/1_et68as0s

PRESENTER: Ben Kelcey, PhD, Assistant Professor of Quantitative Research Methodologies, College of Education, Criminal Justice & Human Services, University of Cincinnati

ABSTRACT: A common approach to investigating the mechanisms through which a treatment is presumed to impact an outcome is a mediation analysis. Mediation analyses examine the extent to which a treatment has an indirect effect on an outcome by examining how changes in a mediator produced by exposure to a treatment come to change outcomes. Despite the critical role mediation analyses have historically played in social science research, literature examining methods for designing and analyzing mediation in cluster or multilevel settings is sparse. In this talk, I discuss recent developments in both the design and analysis of studies of multilevel mediation.

 


 

April 5, 2013

TITLE: Representing Time in Longitudinal Research: Assessment Lag as Moderator

video opens new window  https://mediaspace.itap.purdue.edu/media/Representing+Time+in+Longitudinal+ResearchA+Assessment+Lag+as+Moderator/1_y52wws21

PRESENTER: Todd Little, PhD, Professor, Department of Psychology, Director, The Center for Research Methods and Data Analysis (CRMDA), Director, Quantitative Training Program, University of Kansas

ABSTRACT: I will discuss various ways that time can be used to evaluate change processes in longitudinal research. Moving away from the classic mantra that behavior is a function of change, I present a less constrained view that change is a function of time. This less constrained view leads to numerous innovations in research design and analysis that can unveil hidden and important developmental processes.

 

TITLE: A Formal Test of Moderated Mediation

video opens new window  https://mediaspace.itap.purdue.edu/media/A+Formal+Test+of+Moderated+Mediation/1_prxqcu7n

PRESENTER: Andrew Hayes, PhD, Associate Professor, School of Communication and Department of Psychology, The Ohio State University

ABSTRACT:  Methods for testing a hypothesis about "moderated mediation" that have become widely used in the behavioral science literature are piecemeal approaches that do not formally test whether the indirect effect in a mediation model is moderated. In this talk I introduce a formal test of moderated mediation based on a hypothesis test or interval estimate of the parameters of a function liking the indirect effect to values of a moderator. Real-data examples are provided for different forms of moderated mediation, and the implementation of the method is illustrated using software such as the PROCESS procedure for SPSS and SAS as well as Mplus.

Purdue University, 610 Purdue Mall, West Lafayette, IN 47907, (765) 494-4600

2016 Purdue University | An equal access/equal opportunity university | Copyright Complaints | Maintained by HDFS

If you have trouble accessing this page because of a disability, please contact HDFS.