New PNAS Letter by Prof. Katrin Auspurg: Billions of regressions ≠ robustness
17.11.2025
In a recent PNAS Letter, Prof. Katrin Auspurg reanalyzes a prominent study to demonstrate why multiverse analyses require thoughtful model specifications. Including unjustified models can vastly overstate uncertainty and obscure robust findings.
The Letter argues that “robustness is better assessed with a few thoughtful models than with billions of regressions” and illustrates how mechanically crossing specifications and selecting by model fit can be misleading—leading to a severe overestimation of uncertainty in the social sciences.
For the full publication, see: https://www.pnas.org/doi/10.1073/pnas.2521917122
