Percentiles: how do you work them out?

[Villagedad]

Hi all

I often read people quoting their DC is in the 96th percentile for VR etc etc., and have often wondered how they are calculated?

Any help would be welcomed as we are trying to calculate these for an appeal.

Our DS's 11-plus raw scores / marks we need the percentiles for are as follows:

VR - 141 (raw score 64 out of 80 correct)
NVR - 141 (raw score 50 out of 72 correct)

Many thanks for your help
Villagedad

[Okanagan]

Have a look here for an explanation of standardised scores. A score of 141 for a typical educational attainment test is in the top 1%. It depends whether your test has been scored against a whole population (e.g. Bucks, Kent) or against the cohort who took the test (e.g. Birmingham, Warwickshire) though - as you can't compare results between the two. However the number scoring full marks for a test which is predictable and tutored for tends to exceed the number you would statistically expect (hence the current move in many places towards tests which are more "tutor proof").

[Villagedad]

Have a look here for an explanation of standardised scores. A score of 141 for a typical educational attainment test is in the top 1%. It depends whether your test has been scored against a whole population (e.g. Bucks, Kent) or against the cohort who took the test (e.g. Birmingham, Warwickshire) though - as you can't compare results between the two. However the number scoring full marks for a test which is predictable and tutored for tends to exceed the number you would statistically expect (hence the current move in many places towards tests which are more "tutor proof").

Many thanks Okanagan for your help

We are based in Kent. Would it be better for us to contact NFER to get an accurate percentile for our DS's scores above?

Kind regards
Villagedad

[Okanagan]

In a normal distribution 141 is 99.68% if you want to be exact. But at the extreme ends of the scale the values aren't statistically accurate - hence the reason that scores above 141 aren't published. And higher scores are clearly possible since the scores you are quoting are nowhere near 100% on raw scores.

In Kent for 2012 entry 76 children scored 418, 98 scored 419, and 589 scored the maximum 420.
For 2013 entry (with an increased maximum of 423 - i.e. 3 x 141) 880 scored over 420 for 2013, including over 660 on the full 423.

Clearly that's a lot more than 1% scoring 141 on every section as there obviously weren't over 66000 children taking the Kent test!

[Etienne]

Kent is (to date) relatively straightforward, and I would have thought appeal panels would normally focus on just the standardised score - and have no problem accepting that 141 is at the 99th percentile.

Incidentally, appeal panels will usually be familiar with percentiles and national standardisation from the evidence of CATs which appellants sometimes put forward.
See the table in B28(d):
http://www.elevenplusexams.co.uk/appeal ... cation#b28" onclick="window.open(this.href);return false;


I'm not aware of any meaningful percentiles for the Bucks 11+, because LA representatives used to say it was standardised for the cohort. KenR, on the other hand, has described it as "a bespoke hybrid approach which results in a test mean (average) of about 111-112 rather than 100".

[Villagedad]

Kent is (to date) relatively straightforward, and I would have thought appeal panels would normally focus on just the standardised score - and have no problem accepting that 141 is at the 99th percentile.

Incidentally, appeal panels will usually be familiar with percentiles and national standardisation from the evidence of CATs which appellants sometimes put forward.
See the table in B28(d):
http://www.elevenplusexams.co.uk/appeal ... cation#b28" onclick="window.open(this.href);return false;


I'm not aware of any meaningful percentiles for the Bucks 11+, because LA representatives used to say it was standardised for the cohort. KenR, on the other hand, has described it as "a bespoke hybrid approach which results in a test mean (average) of about 111-112 rather than 100".

Many thanks

Would it be possible (or correct) to quote a percentile figure for a combined Kent score (VR, NVR & Maths)? If so in our case we have a combined score of 400 out of a maximum 423.

Many thanks
Villagedad

[Okanagan]

You can't really aggregate standardised scores and then expect to just divide by 3 to get the equivalent of one, and then give a percentile, as there will be an element of "regression towards the mean" - but if you could 133/141 would be about 98.68%.

You might be able to ask Kent admissions if they could give a figure for what percentage of candidates (or even x out of y candidates) scored an aggregate of 400 or above.