ACT/SAT vs. 2015 MCAT correlation

This forum made possible through the generous support of SDN members, donors, and sponsors. Thank you.
huh. my bad. That's weird though.

By definition 100 percentile would mean that all other test-takers got a lower score (i.e. it should be reserved for those that got a 528). But in this case, statistically speaking, only about 500 test takers will be in the 100 percentile (0.5% of 100k test takers per year). In addition, this means that anyone getting an equivalent of a 39+ gets lumped in that category... which makes sense as the difference between a 39 and 41 can be 2-3 questions.
 
By definition 100 percentile would mean that all other test-takers got a lower score (i.e. it should be reserved for those that got a 528). But in this case, statistically speaking, only about 500 test takers will be in the 100 percentile (0.5% of 100k test takers per year). In addition, this means that anyone getting an equivalent of a 39+ gets lumped in that category... which makes sense as the difference between a 39 and 41 can be 2-3 questions.
The 100th percentile means the 99.5+th percentile
 
By definition 100 percentile would mean that all other test-takers got a lower score (i.e. it should be reserved for those that got a 528). But in this case, statistically speaking, only about 500 test takers will be in the 100 percentile (0.5% of 100k test takers per year). In addition, this means that anyone getting an equivalent of a 39+ gets lumped in that category... which makes sense as the difference between a 39 and 41 can be 2-3 questions.

It'd be more accurate to use 99.9th %ile, like the ACT does. Not that it matters haha
 
By definition 100 percentile would mean that all other test-takers got a lower score (i.e. it should be reserved for those that got a 528). But in this case, statistically speaking, only about 500 test takers will be in the 100 percentile (0.5% of 100k test takers per year). In addition, this means that anyone getting an equivalent of a 39+ gets lumped in that category... which makes sense as the difference between a 39 and 41 can be 2-3 questions.

The point @Axes was trying to make is that usually, 100th percentile isn't actually a thing. That would mean a test taker did better than everyone... including himself/herself, which doesn't make sense. However, AAMC rounds to the nearest whole number, so, as @walloobi pointed out, 100th% means 99.5+%.
 
All right guys I'm sorry lol I don't know! But my score report says "100 %" for my overall percentile rank. I apologizeeeeeee I did not mean to start such a debate over this.
 
It'd be more accurate to use 99.9th %ile, like the ACT does. Not that it matters haha

Agreed. Would make the stats a lot more meaningful/useful for the top scores that are all 100%.
 
Thanks, but I can only use MCAT 2015 scores. Nice improvement though.

As a teacher, I see students everyday who could change the world if they weren't targeted for prison instead. Had they access to authentic resources, you would see a nice improvement in them too -- intelligence is not the limiting reagent.
 
Last edited:
As a teacher, I see students everyday who could change the world if they weren't targeted for prison instead. Had they access to authentic resources, you would see a nice improvement in them too -- intelligence is not the limiting reagent.

I hope this is true. I'll be taking my MCAT two summers from now and for someone who got a 30 on ACT (27 first time, 30 on retake) I am afraid that I'll not be able to score a 512-514 (32-33 old MCAT) when I take it.
 
I hope this is true. I'll be taking my MCAT two summers from now and for someone who got a 30 on ACT (27 first time, 30 on retake) I am afraid that I'll not be able to score a 512-514 (32-33 old MCAT) when I take it.

Depends. No one can tell you. This correlations chart is kind of showing that there is no correlation to ACT and MCAT.
 
Depends. No one can tell you. This correlations chart is kind of showing that there is no correlation to ACT and MCAT.

I guess you can look at it however you want. I see a direct correlation, with a score of 29-30 predicting a score between 508 and 512-514 at most and higher scores on ACT correlating with higher MCAT scores.
 
I guess you can look at it however you want. I see a direct correlation, with a score of 29-30 predicting a score between 508 and 512-514 at most and higher scores on ACT correlating with higher MCAT scores.

I wonder how much of that actually is attributed to the person doing well on tests no matter what? Or if its actually that doing well on the ACT specifically is a correlation to the MCAT.
 
Base on my experience, people who come from middle class+ families tend to have a direct correlation between post high-school examination and the MCAT. Peoples from low SES are more wild cards.
 
Base on my experience, people who come from middle class+ families tend to have a direct correlation between post high-school examination and the MCAT. Peoples from low SES are more wild cards.

That will always hold true in every facet of life though, not just standardized tests.
 
Just curious, what are you getting on practice tests? Have a score that you are aiming for?
I once took the old MCAT and got a 17 only having gen chem and gen bio, no physics or orgo. I have not taken any new practice tests, but will update once I do.
I want a score that will make up for my horrid GPA! :sorry:
 
I once took the old MCAT and got a 17 only having gen chem and gen bio, no physics or orgo. I have not taken any new practice tests, but will update once I do.
I want a score that will make up for my horrid GPA! :sorry:

Best of luck to you!
 
Different results using percentiles rather than the actual numbers?
 
Different results using percentiles rather than the actual numbers?

It should be a similar correlation since higher score = higher percentile.

But the absolute percentiles would differ for each exam. A 30 ACT is 95th percentile and its correlate MCAT (from the graph) is about 509-512, which is 80-85th percentile. This is expected since the test-taking populations differ.
 
Yo, check your homoscedasticity. R^2 scores can be overestimated if there is too little homoscedasticity.
 
Yo, check your homoscedasticity. R^2 scores can be overestimated if there is too little homoscedasticity.

I have been using R values. Not R^2. I have limited knowledge on statistics; it is not a course I have taken yet. Is the coefficient of determination (R^2) the proper measurement for this? How do I measure homoscedasticity? Please expound.
 
Okay. R values and R^2(R squared) are numbers that measure correlation coefficient. There are differences between the two, but R values seem fine for your endeavor.

Now homoscedasticity is a topic that isn't dealt in intro stats classes(at least not in my case). Pretty much homoscedasticity means that your data points are more or less equidistant from your regression line. If you want to measure homoscedasticity, you can use a variety of tests like Breusch-Pagan or Goldfeld-Quandt. However, stuff like dealing with homoscedasticity is something best done by people with extensive backgrounds in stats.

For an internet forum your methods are fine, but if you ever need help tackling homoscedasticity you can PM me your data set and I'll see what I can do.
 
Top