Webinar: Analyzing the 2016 General Social Survey

Overview

In this webinar, we explore the high-dimensional relationships between hundreds of variables from the 2016 General Social Survey.¹ Our objective is to employ one of BayesiaLab's Unsupervised Learning algorithms to generate a Bayesian network model that approximates the joint probability distribution of the underlying survey responses.

Once we have this representation, we can address research questions that specifically require computing the joint probability. In this context, a prototypical—albeit theoretical—issue would be how to best approximate the diversity of many voters through a small number of elected representatives. This translates directly into a variable and data clustering task, which will be at the core of our presentation. Our objective will be to trade off the faithfulness of voter representation with the number of variable/data clusters to be generated.

Workflow

• Transform original GSS data in SPSS format with Stat/Transfer 14.
• Import data, variable labels, and value labels into BayesiaLab.
• Define variable classes, e.g., demographics, voting behavior, etc.
• Perform Unsupervised Learning.
• Optimize the Bayesian network model with Structural Coefficient Analysis.
• Identify Latent Factors through BayesiaLab's Variable Clustering.
• Cluster voters into segments with Data Clustering (Multi-Net).

Presentation Video

References

¹ Smith, Tom W, Peter Marsden, Michael Hout, and Jibum Kim. General Social Surveys, 1972-2014 [machine-readable data file] /Principal Investigator, Tom W. Smith; Co-Principal Investigator, Peter V. Marsden; Co-Principal Investigator, Michael Hout; Sponsored by National Science Foundation. --NORC ed.-- Chicago: NORC at the University of Chicago [producer]; Storrs, CT: The Roper Center for Public Opinion Research, University of Connecticut [distributor], 2015.

1 data file (57,061 logical records) + 1 codebook (3,567p.). — (National Data Program for the Social Sciences, No. 22).