Chances of interview/accepted at mcmaster med

[toronto_sam]

Hey Guys, 

So I was bored a couple days ago and I made this Excel file that calculates your chances to get interviewed and accepted. This was just for fun so take it with a grain of salt. I provided an explanation below on how I did it. I know that somebody posted something similar earlier. But as a math nerd, I realized that there was a mistake - GPA isn't normally distributed (i.e. no zscores, and other stuff). So I used an exponential regression.

TL;DR - Fill in the yellow boxes, and the green gives you the chances

GPA Calculation - take the amount of students who got in with that GPA (take mid-point so 56 students for 3.85, 93 students for 3.95, etc.). Then find the probability - P = Amount in that section and lower/Total #of students. Then plot these (it's exponential!!!) and use a exponential regression to find the value for the input GPA.

MCAT Calculation - find zscore (Z = (Score - Mean)/SD) then corresponding probability (this is normally distributed so we cool)

CASPer - Assumed 5 was mean and 2 was the SD and found the zscore, and probability like MCAT. It's a "how you feel" kinda thing so don't sweat it too much.

MMI - Same as CASPer

Degree - They give you a boost if you have masters/phd so I considered that

Interview Chance = (0.32*GPA + 0.32*CASPer + 0.32*CARS + DegreeBoost)*ScaleFactor

ScaleFactor was chosen to be 1.5. The stats were for people who were ACCEPTED. Since they interview more than they accept, your chances should be a little more for the interview. Also, 1.5 is arbitrary, and a conservative estimate - feel free to change it to whatever lets you sleep at night.

Accepted Chance = 0.15*CARS + 0.15*GPA + 0.7*MMI

Also, these are truncated at 20% and 95% respectively. So below 20% becomes a 0, and above 95% stays at 95. 

I'm sure this could be better, so let me know what y'all think! 

McMaster Chances.xlsx

[Jess85]

This is cool!

When I was applying, I used RUC Score for McMaster to predict my chances of getting an interview.  It's free. 

[indefatigable]

On June 19, 2017 at 1:26 PM, toronto_sam said:

Hey Guys, 

GPA Calculation - take the amount of students who got in with that GPA (take mid-point so 56 students for 3.85, 93 students for 3.95, etc.). Then find the probability - P = Amount in that section and lower/Total #of students. Then plot these (it's exponential!!!) and use a exponential regression to find the value for the input GPA.

 

McMaster Chances.xlsx

This is a very interesting calculation.  I've noticed the same thing for GPA in individual courses - that it's often more exponential like.  

What I find interesting in well, is that in Quebec, admission to most professional school (law, med, dentistry..) is based on a collegial R-score, which is explicitly a modified Z-score from the individual courses.  But if the courses aren't normally distributed (and actually more exponential), this is really going to make the calculations fairly inaccurate (small differences with respect to the mean will be exaggerated).  Theoretically, the course GPAs would only be normally distributed if there were a very large number of examinations in each course OR policies were in place to ensure normality across classes.   

I believe in this era of big data, relying on a single number summary from a class is highly inaccurate.  GPA as it stands is a 19th century US college invention and it's time to at least keep the distributions, since what's being looked for is the academic strength relative to peers rather than the actual number (especially given the lack of standardization across faculties, universities and even individual classes).  But this only solves part of the problem, since different peer groups may be different academically.  

 

[toronto_sam]

@calcanI think you make a fair point. However, I always believed that course marks are inherently normally distributed - the mean and standard deviation may vary but the overall pattern shouldn't as long as the class is big enough (which most are). The problem that I can see is that only people with higher marks actually apply to medical school whereas the ones with lower marks don't even try. I think this is the root cause of the exponential pattern that we see - more people with 3.9+ apply than people with 3.7+. I don't know what sort of policies you can have to solve this problem :/

[indefatigable]

On July 4, 2017 at 1:06 PM, toronto_sam said:

@calcanI think you make a fair point. However, I always believed that course marks are inherently normally distributed - the mean and standard deviation may vary but the overall pattern shouldn't as long as the class is big enough (which most are). The problem that I can see is that only people with higher marks actually apply to medical school whereas the ones with lower marks don't even try. I think this is the root cause of the exponential pattern that we see - more people with 3.9+ apply than people with 3.7+. I don't know what sort of policies you can have to solve this problem :/

@toronto_sam Actually by statistical theory, it doesn't matter how large the classes are only the number of exams and tests (and any grading curve policies in place).  If there are a very large number of exams and tests, the individual grade will be drawn from a normal distribution and otherwise not necessarily so.  I've looked at grade distributions for even large classes ~200 people and they are often exponentially decaying (not close to normal).  It's true that the distribution won't necessarily take shape until there a large number of data points, though. 

In terms of applicants, I believe that at UOttawa for example, at least half the applicants didn't make the wGPA cutoff in the past.  So GPA is typically a big obstacle for many people.  I would guess that at Mac, the lack of strict GPA requirements mean that there are very many people who apply, but a lower GPA (say under 3.6) means that they wouldn't have a great chance at getting interview.  I wonder if the lack of independence in the variables i.e. accepted with lower GPA means must have had great CASPER and/or CARS isn't fully accounted for in the fitting  - but  I'd suppose one would need the data from McMaster to see this.  

[Nimbles]

On 6/19/2017 at 1:26 PM, toronto_sam said:

Hey Guys, 

So I was bored a couple days ago and I made this Excel file that calculates your chances to get interviewed and accepted. This was just for fun so take it with a grain of salt. I provided an explanation below on how I did it. I know that somebody posted something similar earlier. But as a math nerd, I realized that there was a mistake - GPA isn't normally distributed (i.e. no zscores, and other stuff). So I used an exponential regression.

TL;DR - Fill in the yellow boxes, and the green gives you the chances

GPA Calculation - take the amount of students who got in with that GPA (take mid-point so 56 students for 3.85, 93 students for 3.95, etc.). Then find the probability - P = Amount in that section and lower/Total #of students. Then plot these (it's exponential!!!) and use a exponential regression to find the value for the input GPA.

MCAT Calculation - find zscore (Z = (Score - Mean)/SD) then corresponding probability (this is normally distributed so we cool)

CASPer - Assumed 5 was mean and 2 was the SD and found the zscore, and probability like MCAT. It's a "how you feel" kinda thing so don't sweat it too much.

MMI - Same as CASPer

Degree - They give you a boost if you have masters/phd so I considered that

Interview Chance = (0.32*GPA + 0.32*CASPer + 0.32*CARS + DegreeBoost)*ScaleFactor

ScaleFactor was chosen to be 1.5. The stats were for people who were ACCEPTED. Since they interview more than they accept, your chances should be a little more for the interview. Also, 1.5 is arbitrary, and a conservative estimate - feel free to change it to whatever lets you sleep at night.

Accepted Chance = 0.15*CARS + 0.15*GPA + 0.7*MMI

Also, these are truncated at 20% and 95% respectively. So below 20% becomes a 0, and above 95% stays at 95. 

I'm sure this could be better, so let me know what y'all think! 

McMaster Chances.xlsx

Thanks for this :-)