Resources
Text analysis with GPT models
Across the social sciences, researchers are increasingly turning to text as a form of data. Yet, quantifying large volumes of text rigorously and precisely is incredibly difficult. These videos are from a workshop I gave at the Center for the Study of Adolescent Risk and Resilience at Duke University in February 2024. In the first video, I presented the results of my work on using generative pre-trained tranformers (GPTs) for automated text analysis (or "content coding"). In the second video, I show viewers how to do this themselves.
The replication crisis and open science
This video explains what is known as the "replication crisis" and how it began. I also discuss some causes of non-replicable research and some solutions to these problems.
Values and objectivity in science
This video discusses the question of whether science should be value-free. I argue that it cannot be, and that scientists should therefore not strive to be "objective" in that sense. Instead, the ideal of objectivity should be understood in terms of openness, transparence, and convergence in the face of public scrutiny of ideas.
The methods of psychological science
This video gives a broad overview of basic concepts from psychological science. It covers how psychologists measure things, and how they test the quality of measurement instruments. It also addresses the concepts of mediation and moderation, within-person versus between-person relationships, and "internal" versus "external" validity.
What is a p-value?
In science, p-values appear all over the place and are incredibly important. But what are they? What do they tell us? This video explains p-values using an intuitive example involving a tea party.
What is a correlation coefficient?
This video provides a simple, conceptual introduction to correlations, specifically "bivariate" correlations and confidence intervals around estimates of correlation coefficients.
What is regression and what is it for?
This video builds on the one above, explaining how regression models allow researchers to examine the unique or independent associations between predictors and an outcome variable.