Importance of applying early: WashU case study

[who_dis_new_phone]

---

Members don't see this ad.

I am bored so I decided to do an analysis of probability of obtaining interview slots vs time. I took numbers off of Washington University's MD/PhD statistics page.

Assumptions: The process of interview selection is random. Each person has an equal probability of successfully obtaining each separate interview slot, which are independently and identically distributed.

Model: Bernoulli distribution of i.i.d. binomial trials: total # of interview slots = ∑qi*(p*ni), where qi is the number of applicants at timepoint "i", ni is the number of interview slots at timepoint "i", and p is the probability of obtaining each of the interview slots. The sum is over a set of 8 discretized timepoints (approximately evenly distributed intervals).

Solving for p gives p=0.0024, the probability of obtaining any of the interview slots. The probability of receiving an interview is then the expected number of interviews given at each timepoint (i.e. qi*(p*ni)) divided by the number of completed applicants at that timepoint. This is shown in the graph below:

Discussion: As mentioned many times before, submitting early is a good way to increase your odds and submitting late can kill you. The modeled data shows that applicants have a 25% (!!!) chance at being interviewed at WashU's MD/PhD program if they submit their application on the first day. This probability decreases by 50% to 13% chance of interview by October 15, with a sharp change in the slope at November 1. This analysis assumes that the populations of applicants are identical across time, which is of course flawed because applicants who are informed and on-the-ball know to submit early.


Any critiques, thoughts, or comments?

 

[Lucca]

I am bored so I decided to do an analysis of probability of obtaining interview slots vs time. I took numbers off of Washington University's MD/PhD statistics page.

Assumptions: The process of interview selection is random. Each person has an equal probability of successfully obtaining each separate interview slot, which are independently and identically distributed.

Model: Bernoulli distribution of i.i.d. binomial trials: total # of interview slots = ∑qi*(p*ni), where qi is the number of applicants at timepoint "i", ni is the number of interview slots at timepoint "i", and p is the probability of obtaining each of the interview slots. The sum is over a set of 8 discretized timepoints (approximately evenly distributed intervals).

Solving for p gives p=0.0024, the probability of obtaining any of the interview slots. The probability of receiving an interview is then the expected number of interviews given at each timepoint (i.e. qi*(p*ni)) divided by the number of completed applicants at that timepoint. This is shown in the graph below:
View attachment 235031



Discussion: As mentioned many times before, submitting early is a good way to increase your odds and submitting late can kill you. The modeled data shows that applicants have a 25% (!!!) chance at being interviewed at WashU's MD/PhD program if they submit their application on the first day. This probability decreases by 50% to 13% chance of interview by October 15, with a sharp change in the slope at November 1. This analysis assumes that the populations of applicants are identical across time, which is of course flawed because applicants who are informed and on-the-ball know to submit early.


Any critiques, thoughts, or comments?

Well the obvious criticism is that this assumes that interviews are basically given out randomly but they aren’t. The 4.0/528 Nature author URM Veteran Rhodes Scholar from Harvard is going to get the interview as long as they are complete before the deadline. The 3.0/500 with no research will never get an interview. Especially at WashU where the MCAT is far more important than say Iowa.

Useful info if you’re in the middle of the pack where getting an interview or not is essentially random.