So You Want to Be a Principal Investigator?

摘要

Always on the look-out for interesting topics to write about, I was especially intrigued by the article by Dijk et al. describing their online software for predicting the probability that a researcher will become a Principal Investigator (PI); the tool can be found at http://www.pipredictor.com/. Their machine-learning approach compiled PubMed data from more than 25 000 scientists. Their statistical analyses suggest that their tool is able to fairly accurately predict who will become a PI and how long it will take. I found the results impressive, fun to examine but also, in some respects, disappointing. How did the authors determine what it takes for you to be successful in academics? Among the over 200 metrics quantified, they examined the importance of first or second authorship, the number of publications, the impact factor of the journals published in, and even considered the influence of "hot" papers that received more than the average number of citations. The good news is that it is not just these parameters that determine the probability of becoming a PI--quality publications, regardless of journal impact factor, do matter! What else contributes to the likelihood of your academic success? The rank of your university is important in determining your future, perhaps suggesting that a paper published in Analytical Chemistry by someone on the faculty at Harvard is worth more than one by someone affiliated with a less prestigious institution. Obviously, many scientists become PIs without high impact publications. Why is this? The authors found that if the number of first-author articles is high enough, this supposedly compensates for fewer high impact publications.