Post-UCAT Discussion 2020

[dotwingz]

 

[pt88]

.... 2910, the pain is real with this one

 

[Sandwich74]

Is there a reason why the 9th decile scores (VR:670, DM:750, QR: 800, AR: 770) Add up to 2990 not, 2920? Some of the others don't add up as well

 

[whys]

Is there a reason why the 9th decile scores (VR:670, DM:750, QR: 800, AR: 770) Add up to 2990 not, 2920? Some of the others don't add up as well

Your overall percentile is not an average of every individual percentile, so getting 9th decile in every subsection may not give you a 90th percentile overall. Your overall percentile is calculated by adding the individual scores of each subsection and ranking that against everyone else. The top nth% will get 99th, etc.

 

[ucatboy]

HAha i literally just did that
View attachment 3932

The issue is that they round up to nearest 10 so the s.d.s are slightly different so when i calculated based off the two scores it was slightly wrong. Your z scores are a little more precise so ill use ur S.D

we literally set it out the exact same omg

ETA: The mean is now: 2527 and sd is 297.269
View attachment 3933

Yeah I agree on the lack of accuracy part, but I think it only plagues the 4th and 6th deciles (too close to the mean, any rounding is likely to have a big effect on the standard deviation calculation). The rest of the values are within the 290-310 range. Interestingly, if you average the 4th decile S.D. (343) and the 6th decile (249) one, you end up close to 300, not sure whether that's supposed to happen or not but the two errors seemingly cancel out.

 

[ucatboy]

Is there a reason why the 9th decile scores (VR:670, DM:750, QR: 800, AR: 770) Add up to 2990 not, 2920? Some of the others don't add up as well

The phenomenon of S.D. of each section added up /= overall S.D. has got something to do with statistics but I can't find an easy to understand source or explanation for it. But the general rule of thumb is X percentile in each section -> greater than X percentile overall.

 

[Aster]

.... 2910, the pain is real with this one

me too man...

 

[ucatboy]

Is there a reason why the 9th decile scores (VR:670, DM:750, QR: 800, AR: 770) Add up to 2990 not, 2920? Some of the others don't add up as well

I've given it a bit of thought, and I think I can explain why the sum of individual S.D.s will always be equal or greater than the overall S.D.:

The only situation where the sum of individiual S.D.s equals the overall S.D. would be if one student achieved 99th percentile for every section (= 99 overall), one student achieved 98th percentile for every section (= 98 overall) etc. etc.

In reality, the diverse nature of the UCAT subtests means that you're unlikely to be this consistent and top the charts in every single section. What actually happens, and what you'll find here on MSO, is that you might be good in some sections and slightly worse in others. So in reality, that 99th percentile VR, DM, QR and AR score doesn't all go to the same person - the 99th percentile VR score might belong to the someone who got 80th percentile in DM, for example.

When this "mixing" of scores occurs, everyone's scores get closer together i.e. instead of student A with 900 in every section and student B with 600 in every section, student A might have 600, 900, 600, 900 and student B 900, 600, 900, 600. This represents a "narrowing" of the normal distribution curve:



As a result, the overall standard deviation has reduced, going below the sum of the individual standard deviations. This is the reason why if you get, let's say, 1.282 standard deviations above the mean (equivalent to 90th percentile) for the individual subtests VR, DM, QR and AR, your actual OVERALL standard deviation above the mean (and hence, overall percentile) is a fair bit higher than that.

 

[Noodlemeister]

.... 2910, the pain is real with this one

Same. Here's hoping that we still have a shot at an interview.

 

[ucatboy]

Just for kicks, I thought I'd upload the bell curve for this year's UCAT results using the most accurate data that we have, but this time accounting for the actual testing population (14,000):

I realise that last time my graph had decimal points on the y-axis, because the area under the curve adds up to 1 ("everyone"). I've adjusted it so that the area under the curve now adds up to 14,000, meaning that you can locate your score on the x-axis and find out approximately how many people share the same score as you on the y-axis.

For example, if you scored 3150, approximately 19 other people got the exact same score in ANZ.

 

[TKAO]

Christ almighty those test statistics are WHACK what the heck....110 points above the mean last year? I mean that would mean that people who got 94th percentiles and below basically scraped or dropped below a 90th this year. This is insane wow. I guess the prediction of covid compressing the UKCAT's grade inflation by 100 points in 5 years was correct

 

[soulja]

how do the reported mso scores compare to official stats?

 

[ucatboy]

how do the reported mso scores compare to official stats?

Like this:

Everything is to scale btw, MSO users represent a not-insignificant proportion of the testing population, especially at the higher end of scores

This should serve as a visual reminder that if you think your scores aren't that great compared to everyone else on MSO, take a step back and look at the bigger picture of everyone sitting the test: chances are that you're well into the upper range of the bell curve.

 

[IQ ™]

Christ almighty those test statistics are WHACK what the heck....110 points above the mean last year? I mean that would mean that people who got 94th percentiles and below basically scraped or dropped below a 90th this year. This is insane wow. I guess the prediction of covid compressing the UKCAT's grade inflation by 100 points in 5 years was correct

I would've gotten 95th percentile based on 2019 results, I've basically scrapped a 91st/92nd percentile this year (sat the UCAT this year for the second time). It's painful when you realise the previous test cycle comes back to haunt you a year later

 

[dotwingz]

I guess they were right when they said 27th of August was going to be a cursed day

People on TikTok are freaking out over the 27th August and now I am too

Honestly, wtf is happening on August 27th?

thetab.com

 

[ucatboy]

Christ almighty those test statistics are WHACK what the heck....110 points above the mean last year? I mean that would mean that people who got 94th percentiles and below basically scraped or dropped below a 90th this year. This is insane wow. I guess the prediction of covid compressing the UKCAT's grade inflation by 100 points in 5 years was correct

I would've gotten 95th percentile based on 2019 results, I've basically scrapped a 91st/92nd percentile this year (sat the UCAT this year for the second time). It's painful when you realise the previous test cycle comes back to haunt you a year later

I don't think it's 100% correct to say stuff like "someone who got x percentile last year would drop by y percentiles this year" or "I would've gotten a higher percentile if I sat the exam last year", because if you really DID sit the exam last, your overall score would probably be lower too, maintaining your percentile. I guess it's no longer a matter of improving your score each time you sit, but improving your score FASTER than everyone else - everyone's improving, and you gotta keep up with the competition to maintain your standing.

 

[IQ ™]

I don't think it's 100% correct to say stuff like "someone who got x percentile last year would drop by y percentiles this year" or "I would've gotten a higher percentile if I sat the exam last year", because if you really DID sit the exam last, your overall score would probably be lower too, maintaining your percentile. I guess it's no longer a matter of improving your score each time you set, but improving your score FASTER than everyone else - everyone's improving, and you gotta keep up with the competition to maintain your standing.

I agree with you there. It's just more about hypothetically comparing your score with different test cycles without taking much significance to it anyway. (Because different test takers, different exam etc)

 

[A1]

this time accounting for the actual testing population (14,000)

Last year's 14,000 was affected by Qld Y12 being half the usual cohort size. You need to do a take-two with 16,000

 

[N182]

improving your score FASTER than everyone else - everyone's improving, and you gotta keep up with the competition to maintain your standing.

Surely 90th won't go up from 2920 next year? So I think this year might just be an exception and you simply have to increase yourself.

 

[cocodreams]

Surely 90th won't go up from 2920 next year? So I think this year might just be an exception and you simply have to increase yourself.

Why might this year be an exception?