Ohio State is in the process of revising websites and program materials to accurately reflect compliance with the law. While this work occurs, language referencing protected class status or other activities prohibited by Ohio Senate Bill 1 may still appear in some places. However, all programs and activities are being administered in compliance with federal and state law.

Quantitative Psychology Brownbag

Shannon Jacoby and Frank Leyva Castro
Mon, April 4, 2022
12:30 pm - 1:30 pm
Psychology Building 035

Please join us next week on 4 April 2022, Monday @ 12:30-1:30pm EDT for our very own Shannon Jacoby and Frank Leyva Castro.
Venue: Psychology 35

Quantitative Psychology Brownbag

Shannon Jacoby
Department of Psychology
The Ohio State University

Title: Nonlinear Relationships: Preserving Information and Increasing Interpretability

Abstract: The goal of this project is to investigate one potential method to improve the interpretability of nonlinear data in psychological research settings. Nonlinear data can be modeled by polynomial regression models, but the interpretation of these models is neither straightforward nor commonly practiced. In an effort to improve interpretability of nonlinear relationships while minimizing information lost due to categorization of a continuous variable, the lowess smoothing function was used to create a well-fitting line for our example data. Multiple polynomial models were then generated to recover the lowess line. The R2 of the polynomial model whose predicted line was identified as most similar to the lowess line was used as an approximation of the information contained in the lowess line. From there, categorical models were constructed through various methods of binning the continuous predictor variable. The results indicate that the fifth-degree polynomial model most closely replicated the lowess line, and its R2 value of .1961 was used to evaluate the bin models. Bin model Theory 4, R2=.1925, was closest to this criterion. For a loss of information in the amount of R2=.0036, or one third of a percent, interpretability of this nonlinear relationship was greatly improved.

Discussant: Frank Leyva Castro

 

Frank Leyva Castro
Department of Psychology
The Ohio State University

Title: An overview of Structural Equation Modeling assumptions

Abstract: Statistical assumptions are an imperative component of many statistical procedures, however there is a limited coverage of assumptions in Structural Equation Modeling . A brief literature review shows inconsistent approaches to assumptions in SEM addressing either causal relationships, estimation assumptions or model assumptions. This talk will focus on assumptions relevant to psychological research using SEM with goals to establish a pedagogical piece that helps practitioners to properly evaluate their procedures.  

Discussant: Shannon Jacoby

More information on research talks