LizzyM calculator percentiles don't seem possible?

[KatsuCurry]

I was scrolling through this page, which gives you a percentile of "Where you fall on the LizzyM distribution curve of all applicants?". *Specifically the 2016-2019 version.

Seems like something is off about the numbers. Here's one thought experiment:

To achieve a LizzyM of 77, you need a minimum of a 520 on the MCAT (37 old-score), since you can't have above a 4.0 GPA. However, according to the page I linked, a LizzyM of 77 is "only" in the top 5% of applicants (roughly top 2,500, given 50,000 applicants). But that doesn't make any sense. A 520 is in the top 2% of all examinees, which given 80,000 MCAT takers a year, is in the top 1600 of test-takers. So at worst, you should be in the top 3% of all applicants (1,600/50,000). But top 3% is probably still high since many 520+ scorers don't have a 4.0, let alone a 3.9. The number of people who apply with old scores shouldn't make a difference, because roughly the same amount of people who have scores from past years will take the MCAT this year but apply in future cycles (the numbers cancel out).

I also take issue with the shape of the graph. How is there a bump at the LizzyM 76-80 range? Both GPA and MCAT should be roughly normally distributed, and as such, the LizzyM graph should also be normally distributed.


Does any of this really matter? Not really. Should I be writing my secondaries instead? Yes. But I'm an economics major who's obsessed with statistics so I wanted to point these issues out.

---

Members don't see this ad.

 

[Isomerase]

I can't speak to the program itself but if you need hard data the GPA/MCAT grid suggests that those with 3.8+/518+ represents the top 5.3% of applicants over the last few years. So yes, LizzyM 77+ probably represents the top 2% of applicants or so. I'm sure the sdn calculator runs on some imperfect estimations. Now go write your secondary's.
https://www.aamc.org/system/files/2020-04/2019_FACTS_Table_A-23_0.pdf

 

[Goro]

I was scrolling through this page, which gives you a percentile of "Where you fall on the LizzyM distribution curve of all applicants?". *Specifically the 2016-2019 version.

Seems like something is off about the numbers. Here's one thought experiment:

To achieve a LizzyM of 77, you need a minimum of a 520 on the MCAT (37 old-score), since you can't have above a 4.0 GPA. However, according to the page I linked, a LizzyM of 77 is "only" in the top 5% of applicants (roughly top 2,500, given 50,000 applicants). But that doesn't make any sense. A 520 is in the top 2% of all examinees, which given 80,000 MCAT takers a year, is in the top 1600 of test-takers. So at worst, you should be in the top 3% of all applicants (1,600/50,000). But top 3% is probably still high since many 520+ scorers don't have a 4.0, let alone a 3.9. The number of people who apply with old scores shouldn't make a difference, because roughly the same amount of people who have scores from past years will take the MCAT this year but apply in future cycles (the numbers cancel out).

I also take issue with the shape of the graph. How is there a bump at the LizzyM 76-80 range? Both GPA and MCAT should be roughly normally distributed, and as such, the LizzyM graph should also be normally distributed.


Does any of this really matter? Not really. Should I be writing my secondaries instead? Yes. But I'm an economics major who's obsessed with statistics so I wanted to point these issues out.

The LM score is merely a rule of thumb.

It's not something that's going to be used in Lancet or JAMA.

 

[NPComplete]

I was scrolling through this page, which gives you a percentile of "Where you fall on the LizzyM distribution curve of all applicants?". *Specifically the 2016-2019 version.

Seems like something is off about the numbers. Here's one thought experiment:

To achieve a LizzyM of 77, you need a minimum of a 520 on the MCAT (37 old-score), since you can't have above a 4.0 GPA. However, according to the page I linked, a LizzyM of 77 is "only" in the top 5% of applicants (roughly top 2,500, given 50,000 applicants). But that doesn't make any sense. A 520 is in the top 2% of all examinees, which given 80,000 MCAT takers a year, is in the top 1600 of test-takers. So at worst, you should be in the top 3% of all applicants (1,600/50,000). But top 3% is probably still high since many 520+ scorers don't have a 4.0, let alone a 3.9. The number of people who apply with old scores shouldn't make a difference, because roughly the same amount of people who have scores from past years will take the MCAT this year but apply in future cycles (the numbers cancel out).

I also take issue with the shape of the graph. How is there a bump at the LizzyM 76-80 range? Both GPA and MCAT should be roughly normally distributed, and as such, the LizzyM graph should also be normally distributed.


Does any of this really matter? Not really. Should I be writing my secondaries instead? Yes. But I'm an economics major who's obsessed with statistics so I wanted to point these issues out.

Interesting points.

The fact of the matter is that GPA and MCAT are not entirely independent events. While there's plenty of lower GPA, high MCAT, and even more high GPA, low MCAT combinations, doing well in your science classes implies a fairly strong foundation in the subject areas tested by the MCAT. Regardless, taking the distribution of an aggregate score combining two normalized statistics does not promise a normalized final curve. Height and weight may be normalized curves, independently, but would BMI be guaranteed to be normalized?

If I recall correctly, the data has not been updated, and it does not include a sample size as full as the previous distribution from 2013-2016. It could have been updated though.

Of course, medical school applicant pools get more and more competitive every year. Gap years and non-traditional pathways are all the more common. One can take the MCAT in 2018, get a 520+, and wait a few application cycles to apply. As seen plastered all over this website, there's certainly no shortage of neurotic, high-scoring applicants who are set on getting into T20 schools, and many of these people are willing to take gap years to do so. Are these people less common than most applicants? Yes, of course, by definition. However, that could explain the increased number of applicants on the right lip of the curve.

Also, to my knowledge, reapplicants are re-represented in the aggregated data. Since 60% of people don't get in, a very strong majority of applications every year stem from reapplicants. This means that the same individual could represent several entries/data points in the curve. Reapplicants may have gone on to retake their MCATs or raise their GPAs, shifting the curve to the right. Some may have improved drastically enough to land them in the LizzyM 75+ range, while also being represented in previous application cycles with lower scores.

Maybe the reasons I mentioned above aren't the reason. But it goes to show that there's a lot of meaning lost behind these aggregated numbers.

 

[deleted555445]

I was scrolling through this page, which gives you a percentile of "Where you fall on the LizzyM distribution curve of all applicants?". *Specifically the 2016-2019 version.

Seems like something is off about the numbers. Here's one thought experiment:

To achieve a LizzyM of 77, you need a minimum of a 520 on the MCAT (37 old-score), since you can't have above a 4.0 GPA. However, according to the page I linked, a LizzyM of 77 is "only" in the top 5% of applicants (roughly top 2,500, given 50,000 applicants). But that doesn't make any sense. A 520 is in the top 2% of all examinees, which given 80,000 MCAT takers a year, is in the top 1600 of test-takers. So at worst, you should be in the top 3% of all applicants (1,600/50,000). But top 3% is probably still high since many 520+ scorers don't have a 4.0, let alone a 3.9. The number of people who apply with old scores shouldn't make a difference, because roughly the same amount of people who have scores from past years will take the MCAT this year but apply in future cycles (the numbers cancel out).

I also take issue with the shape of the graph. How is there a bump at the LizzyM 76-80 range? Both GPA and MCAT should be roughly normally distributed, and as such, the LizzyM graph should also be normally distributed.


Does any of this really matter? Not really. Should I be writing my secondaries instead? Yes. But I'm an economics major who's obsessed with statistics so I wanted to point these issues out.

its based off linear assumptions throughout pretty wide bins on a data set that contains 2 of about 10 attributes that a candidate is evaluated on, so its not going to be super accurate. HOWEVER, if you look at this chart, you can see that the average MCAT for an applicant is significantly higher than the average MCAT for a test taker (average for all MCAT takers is around 501). If you assume a normal distribution using this standard deviation, that puts a 520 MCAT as ~93rd %ile for applicants. given there is a decent correlation between MCAT score and GPA, and that small changes in GPA aren't weighted as highly as small changes in MCAT (i.e 3.95/524 is worth more), I could see 4.0/520 as being ONLY the 95th %ile of applicants in terms of stats.