PI Probability

As problematic as a career in science can be, the endgame for most is still described as becoming a principal investigator (PI). Generally speaking, this is the same thing as a tenured track faculty member of a university. With that in mind a couple of guys decided to try and generate a prediction algorithm for one's likelihood of becoming a PI. You can check it out here. It is based on machine learning from publication and author records procured from PubMed. To summarize it focuses primarily on a publication record, taking into account the number of publications, the "impressiveness" of the publications' journals, and the number of first author publications, but it also incorporates gender (with women less likely to become PIs) and the length of your career (judged primarily on the date of your first publication). Whether or not these are good predictors or not is debatable, but this could be a useful tool so I went ahead and gave it a try. Here are my results: 

Wow! A greater than 50% chance? Right off the bat, you gotta be a little bit suspicious of this algorithm if it is giving me basically a coin flip's chance of becoming a PI. Thats far too generous, but I really like that I score so well (90%) amongst all those that are not a PI. I think this means I am rather PI-like...amongst those who never become PIs. Depending on how you look at that its either like a pitcher who somehow leads the team in RBIs or its like being given an especially shiny silver medal. Either way, its a true accomplishment. Now, years ago I realized that the life of PI was probably not for me, but having such an unexpectedly decent chance of becoming a PI forced me to reconsider whether I would ever want to be a PI. Here are my top reasons for being a PI:

• Generally you get your own office.
• There would be a firm "no journal club" rule in my lab. Journal club is easily the least beneficial meeting that is mostly commonplace in science. Has anyone ever looked forward to journal club? No. How many people are pleased if journal club is ever canceled? Everyone.
• Would really celebrate the accomplishments and publications of lab members so as to make these supposed worthwhile events actually seem maybe sort of worthwhile.
• I think you get to pawn off most your work to others. At least all the research, data collection, and analysis.
•  Lab meetings would not be held if they were simply to provide for a discussion between the PI and the individual(s) presenting, thus wasting the time of everyone else in attendance.
• At the very least, Fridays would be lab lunch days, but lunch outings on any day would be welcomed.
• I bet I'd get a lot more emails. I like emails.
• In the event that something important was happening, like the World Cup, the lab would accommodate the viewing of such important happenings, rather than shaming people into watching over their shoulders while viewing the events on their own.
• Explaining that I am a professor would be infinitely more understandable to those not familiar with science careers.
• Probably would actually get to wear some baller PhD robes from time to time at graduations. 
• Being a professor would really help with the distinguished aura I've been trying to cultivate.

Considering that none of the reasons listed have much to do with pursuing independent projects of interest or devoting all your time to reading and writing papers/grants, it is probably a good thing that I am not really gunning for a PI position. Thats probably fine since there are other statistics that suggest that only about 10% or so of people receiving a PhD will go on to become PIs and the number of people receiving PI positions has steadily been decreasing over the last couple of decades. How could that be if even I was given a 57% chance of becoming a PI? In fact, almost all PI predictions that I have seen reported from this algorithm are greater than 50% so something just isn't adding up, because nowhere near half of PI position applicants are having such success. There could be a reporting bias or perhaps despite great odds of success, most PhDs are just not applying for PI positions. That seems unlikely. I think the truth is probably closer to the fact that, like most scientific tools, this one also suffers from uncertain amounts of sensitivity and specificity (in other words, it doesn't work too well). Yes, becoming a PI still seems to take a few parts hard works, a bit of who you know, and just a dash (or several hundreds) of good ol' fashion luck.