[Editor's note: This article, and the mathematical model used in the app below, has been modified since published on 2 June. The article was modified to include more explanation about the model, and the model was updated to make its predictions more accurate.]
Today in Current Biology, three early-career scientists—two postdocs and a new principal investigator (PI)—published a paper that uses data scraped from PubMed to calculate the odds of becoming a PI. David van Dijk, the article's first author, and this correspondent built an app that allows you to input your data and calculate your own odds. (See our Q&A interview with the authors and related coverage at Science.)
With equal credentials, women’s probability trails that of men—always.
The authors of the article devised a short list of factors that appear to have an outsized impact on your career progression: h-index; first-author publications, citations of your most-cited first-author paper; total number of publications; the impact factor (IF) of your highest-impact journal; and the world ranking of the university you work at. All of this depends on your seniority, as measured by the number of years that have passed since your first, first-author publication, so that’s included, too. (This is a simplified version of the model. For the full model, visit pipredictor.com.)
What does it mean to be a PI? The authors figured that if you’re in the last-author spot, you’re probably in charge of the project, so they used that—last-author position on a paper—as a proxy for PI-ship. In playing with the numbers—see below—it quickly becomes apparent that, measured in this way, PI-ship is a much easier thing to achieve than, say, a faculty position at a research university—which only a small fraction (perhaps 20%) of all Ph.D. scientists ever achieve.
Before inviting you to play around with the app yourself, we will point to a few results you can look for:
- With equal credentials, women’s probability trails that of men—always.
- First-author publications are among the most important predictors of career success (as measured by the probability of becoming a PI)—but middle-author publications are less helpful.
- Those who believe that journal IF should not be used in hiring will not be happy with the strong dependence the authors found on the IF of the highest-rated journal you’ve published in.
[After being alerted to differences between our widget’s prediction and that of the full model, Science consulted with the study's authors, who discovered problems with the math they had used with the simplified model—but not, we emphasize, with the full model used in the Current Biology article. The math was revised, and at roughly 10 a.m. on 3 June, the model was updated to reflect the new math. Still, the authors—including lead author David van Dijk—emphasize that the model used here is a simplified version of the model presented in Current Biology and on pipredictor.com. Here's how van Dijk puts it: "The full model has an 'area under curve'—a measure of the model's predictive power—of 0.83 while the six-feature model presented here has an AUC of 0.75. (One is perfect.) As a consequence, the simplified model is less accurate. Still, the calculated probabilities should be roughly correct, and the dependence of the probability on the metrics included is reliable."]
A quick technical note before you dive in: Unsurprisingly, the authors found that the ranking of your current university is important. The ranking they used is the Academic Ranking of World Universities, which you can find here.