2019 Ranking

[AK loving Dad]

What was the breakdown for each area AK loving dad?

Thanks

It was 110.5 VR and 113.5 for NVR Maths

I can find out when I get home the split between NVR Maths

[MSD]

Firstly, let's assume the standard deviation being used for this years test is 15. Educational establishments tend to use between 12.5 and 15 and deviation can vary for each test. A low standard deviation indicates that the data points tend to be close to the mean (also called the expected value, in this case 100 for each area) of the set, while a high standard deviation indicates that the data points are spread out over a wider range of values. So, for an easier test, you will find the standard deviation to be set at a lower value than a difficult test.

Verbal Reasoning Z-number

Z = (110.5-100)/15 = 0.7

Z-number of 0.7 gives us 0.758, which equates to approximately 24.2% scoring higher for that subject area. Approximate Rank Order out of 3369 children - 815

Numeric/NVR Z-number

Z = (113.5-100)/15 = 0.9

A Z of 0.9 gives us 0.8159, which equates to approximately 18.4% scoring higher for that subject area. Approximate Rank Order out of 3369 children - 620

It's important to note the consortium must also age standardise on top of the normal SD numbers, and also combine the above 2 subject areas to produce an aggregate rank. And, please also note, on a score like 224 you can have many candidates, 32 to be exact in the link that was uploaded here, and either side of 224 there is a mysterious missing gap of around 35 ranks.

I understand your DS was ranked 800th out of 3369 and keeping in mind number of candidates on each SD point close to 224, that does look pretty close to the actual rank based on our calculation. Just a point or two away, if we worked on the basis of 224 rather than each area - and we have to account for varying standard deviation.

Hope that makes some sense

Quite a lot of variants involved, unfortunately. This is just a simplified tool to work out an approximate rank based on all the assumptions we have made so far.

[AK loving Dad]

Firstly, let's assume the standard deviation being used is 15. Educational establishments tend to use between 12.5 and 15 and deviation can vary for each test. A low standard deviation indicates that the data points tend to be close to the mean (also called the expected value, in this case 100 for each area) of the set, while a high standard deviation indicates that the data points are spread out over a wider range of values. So, for an easier test, you will find the standard deviation to be set at a lower value than a difficult test.

Verbal Reasoning Z-number

Z = (110.5-100)/15 = 0.7

Z-number of 0.7 gives us 0.758, which equates to approximately 24.2% scoring higher for that subject area. Approximate Rank Order out of 3369 children - 815

Numeric/NVR Z-number

Z = (113.5-100)/15 = 0.9

A Z of 1.866 gives us 0.8159, which equates to approximately 18.4% scoring higher for that subject area. Approximate Rank Order out of 3369 children - 620

It's important to note the consortium must also age standardise on top of the normal SD numbers, and also combine the above 2 subject areas to produce an aggregate rank. And, please also note, on a score like 224 you can have many candidates, 32 to be exact in the link that was uploaded here, and either side of 224 there is a mysterious missing gap of around 35 ranks.

I understand your DS was ranked 800th out of 3369 and keeping in mind number of candidates on each SD point close to 224, that does look pretty close to the actual rank based on our calculation. Just a point or two away, if we worked on the basis of 224 rather than each area - and we have to account for varying standard deviation.

Hope that makes some sense

Quite a lot of variants involved, unfortunately. This is just a simplified tool to work out an approximate rank based on all the assumptions we have made so far.

Thank you - much appreciated - you have helped many of us on many topics.

[MLW]

We registered in Warwickshire and we were informed that the maximum point difference between a September born child and a July child would only be up to 3 points.

[KenR]

We registered in Warwickshire and we were informed that the maximum point difference between a September born child and a July child would only be up to 3 points.

Three points for each standardisation or three point in total? Suspect the former

[MLW]

https://apps.warwickshire.gov.uk/api/do ... CC-699-938" onclick="window.open(this.href);return false;
Sorry I made an error, the difference between a September born child and an August born is ‘on average’ a total of 4 points.

[KenR]

https://apps.warwickshire.gov.uk/api/do ... CC-699-938
Sorry I made an error, the difference between a September born child and an August born is ‘on average’ a total of 4 points.

If you read the addendum wording carefully you will see that they are not saying the difference is 4 for this test, but typically an average of 4 for most 11+ tests (viz. across the country)

[MLW]

https://apps.warwickshire.gov.uk/api/do ... CC-699-938
Sorry I made an error, the difference between a September born child and an August born is ‘on average’ a total of 4 points.

If you read the addendum wording carefully you will see that they are not saying the difference is 4 for this test, but typically an average of 4 for most 11+ tests (viz. across the country)

Correct, but this 11 plus test will be included in that generic statement. It is therefore unlikely that there is a overall difference of 12 points in the test.

[KenR]

Correct, but this 11 plus test will be included in that generic statement. It is therefore unlikely that there is a overall difference of 12 points in the test.

Actually, that's not correct for CEM exams. I did obtain the actual standardisation tables from the FO for the first couple of CEM exams in 2004 and 2005. (They won't issue them now as they claim it is CEM copyright and confidential)

In those days they used to standardise the VR, NVR and Maths/Numerical separately and add together to give a total standardised score of typically 300+

For example, in the Birmingham KE VR Tests in Nov 2004 , to achieve a Standardised Score of 118(the average min score per section to get into KECHB) a child aged 11.18 years on the exam day would have to achieve score 71/100, whereas a child who was only 10.22 years would only have to score 65/100. Quite a difference.

The range were similar for NVR and Numerical sections as well.

So for this particular exam the difference could be a high as 18 for children at the extremes of age on the day of the exam. As there are only 2 accumulative scores in the current exam the difference won't be as high as 18 but based on actual empirical data it certainly could be as high as 12.

Hope this helps

[MSD]

I have provided an explanation of age standardisation before and will try again using a very simplistic approach.

When undertaking the standardisation of scores, CEM only compare students of the same age. CEM draws up 12 age groups of students starting from Sept(Oldest) to August(Youngest).

So, all the September born children will undergo the standard standardisation process used by CEM, and get their standardised scores, separately. Then the same process will happen for the October born children, etc. In each birth month group, the student scoring average amongst that group will get a standardised score of 100 and each student scoring above average will get a 100+ score based on standard deviation used and their actual raw score.

What the above analysis shows is that you DO NOT have to be in the top X percentage of all cohort taking test to have a chance of a place, it means that YOU HAVE to be in the top X percentage of all those having their birthday in the same month as you. So, a September child has to be in the top X of all September children, an August child has to be in the top X of all the August children, etc. to get the standardised score which will get you a high enough ranking and cut-off to gain entry. And you will find the X will be between 20 and 25% for Birmingham Grammar schools.

So, hypothetically, if you had a really clever younger month group in a certain test, you may find their SD score is in fact higher for the same raw score achieved as compared to an older month group. Hence, age standardisation is based on the test and what each month group scores and what is their average. For an easier test, you may find the average across all age groups is pretty similar and age standardisation doesn't really come into play.