There are many ways of conducting an analysis, but most studies show only a few carefully curated estimates. Applied research involves a complex array of analytical decisions, often leading to a ‘garden of forking paths’ where each choice can lead to different results.
By systematically exploring how alternative analytical choices affect the findings, Multiverse Analysis reveals the full range of estimates that the data can support and uncovers insights that single-path analyses often miss.
Go beyond single-path methods and discover how multiverse analysis can lead to more transparent, illuminating, and credible contributions to science
Multiverse Software: In Stata, type “ssc describe multivrs“
Related Articles
“Model Uncertainty in Sociological Research: An Application to Religion and Economic Growth.” American Sociological Review. June 2009.
“We Ran 9 Billion Regressions: Eliminating False Positives Through Computational Model Robustness.” 2018. With John Muñoz. Sociological Methodology. Vol 48(1): 1-33. — Symposium article, with commentary by Bruce Western and Robert O’Brien.
Download the replication package
“Rejoinder: Can We Weight Models by Their Probability of Being True?” 2018. With John Muñoz. Sociological Methodology. Vol 48(1): 43–51.
“Model Uncertainty and the Crisis in Science.” 2018. Socius.
“Model Uncertainty and Robustness: A Computational Framework for Multi-Model Analysis.” Sociological Methods and Research. February 2017. Vol. 46(1): 3 – 40. (Lead article, with Katherine Holsteen)
Download: This do file installs our Stata program, loads in data sets, and replicates all the analyses in the paper.
OR: type “ssc describe multivrs” in Stata
The multiverse framework allows researchers to
(1) estimate thousands of regression models across combinations of control variables, functional forms, alternative variable definitions, and estimation commands;
(2) display the resulting modeling distribution of estimates; and
(3) discover which aspects of model specification have the most influence on an estimate of interest.
This allows analysts to clarify and demonstrate which modeling assumptions are essential to their empirical findings, and which are not.