Skip to content
 

My talk this Wednesday at Stanford business school

It’s in the Organizational Behavior Seminar, Wed 7 Mar at noon in room E247:

Toward replicable research in the human sciences: How can we get from where we are, to where we want to be?

We’ve heard a lot about the replication crisis in science. Now it’s time to consider solutions from several directions including institutions (changing incentives in scientific practice and reporting), design and data collection (higher-quality measurement and within-person designs), and statistical analysis (multilevel modeling and Bayesian inference or regularization). We discuss examples from political science, economics, psychology, and medicine.

We also discuss some widely recommended ideas which we think would be useless or even counterproductive, ideas which are not as rigorous as one might think.

Here are some relevant articles:
http://www.stat.columbia.edu/~gelman/research/published/incrementalism_3.pdf
http://www.stat.columbia.edu/~gelman/research/published/jasa_signif_2.pdf
http://www.stat.columbia.edu/~gelman/research/unpublished/abandon.pdf

I’ll try to avoid repeating any material from my previous talk at the Stanford business school, two years ago.

If they ever invite me a third time, I’ll forget about social science entirely and just give them some straight Bayes: our recent research in partial pooling, informative priors, stacking, model evaluation, and all the rest. Actually, I will talk about some of that stuff in my talk this Wednesday.

P.S. It looks like the talk is in a small room, so if you want to attend, you should contact the department ahead of time at the above link, I guess.

3 Comments

  1. Kyle C says:

    The statistics course I took there in the mid-90s literally taught us R-hacking. I hope things have improved.

  2. Guive says:

    Any chance you’ll be in Berkeley while you’re on the West coast?

Leave a Reply