USMLE Looking for feedback on predicting your step score

Started by Laradd
This forum made possible through the generous support of SDN members, donors, and sponsors. Thank you.

Laradd

Full Member
5+ Year Member
Advertisement - Members don't see this ad
With all this covid pandemic forcing me to be home I wanted to work on something useful for my fellow colleagues. I remember when I was taking my step exam that I was really frustrated with trying to estimate what my score would be with the practice tests and digging through posts on student doctor and reddit to try and find a correlation. I've spent the past week gathering up all the data I could find on student doctor, reddit and other forums and putting it all together into one giant algorithm. I present to you the fruits of my labor: predict my step score

It's still a work in progress so if anybody has any feedback or suggestions it would really be appreciated. Also if you have already taken your exam you can still submit your score which i use to improve peoples predictions in the future so that is always appreciated.
 
My question is why are there two boxes? One says 230 in gray and the other says 30 without actually putting anything in. What are you supposed to put in the second box on the right? Thanks for making this btw! Also how accurate is this? It says p value <.005 which is appears accurate but it puts me at a 246 and there is no way in heck thats accurate with a 217 baseline NBME 21 and 73% first pass UW% of which im only 50% through on timed random specially since my exam is in June.
 
Last edited:
It has instructions right at the top and you probably set the days before as your exam score so if you got a 217 on NBME 21 and you took it 217 days before it is very likely you'll improve to 246 in 217 days lol
 
I helped out with this project as well transferring alot of the data on sdn and it's just based on the weights associated with those exams so its technically feasible. Unfortunately NBME 21 is one of those weird exams where I wasn't able to get alot of data. On top of that, according to the data we have, 21 underpredicts so that is part of it as well
 
Last edited: