Standardisation
Moderators: Section Moderators, Forum Moderators
I am most definitely not a teacher basher!Ed's mum wrote:Oh deep joy, more potential teacher bashing!! What a novelty... yawn
IMHO teaching is a vocation. If wages/social standing etc were improved it would attract more educated people. This is not a personal comment!.
To answer the OP's post again, I would add that as the mother of a child with a November birthday I was aware that she would need to score higher than those who were summer born. Age standardisation simply negates the effects of immaturity even when the tests are partially based on the NC.
-
- Posts: 1245
- Joined: Fri Jul 06, 2007 9:31 pm
Re: Understanding standardised scores
That's nice. It proves they can do well in 11+ tests at least.mattsurf wrote:On the Warks pages one person said that their child came 1st in the exam with scores of 143,143,139 (avg. 142).
The numbers are normalised to have mean 100 and std deviation 15. Theoretically there are small numbers of samples that could be far up and down the tails. I think this is why normally the results get capped at 69/70 & 140/141 rather than reporting numbers such as 143. But you learn something everyday!mattsurf wrote: Any idea what the upper and lower limits for the standardisation curve would be here and how the scores would be distributed on the bell curve?
If you want to talk percentiles you need to look at the area under the curve.mattsurf wrote: ...this is a statistically the 99.99th percentile
The top 2% will have a standardised score of 130 or above.
The top 5% will have a standardised score of 125 or above.
The top 10% will have a standardised score of 120 or above.
The top 30% will have a standardised score of 108 or above.
The top 50% will have a standardised score of 100 or above.
The top scorer is by definition in the top bucket!
Conversely 50% of the sample will get scores of less than 100. But in a self-selecting group they could still be VERY smart compared to the overall population.
Regards
SVE
Animis opibusque parati
Warks is now a CEM exam; CEM exams are often (but not always) standardised around 100/15 SD but not capped at either end, so scores of 143 (or presumably zero) are entirely possible. In Birmingham there have been cases of people scoring more than 500 over the three categories in the KE Foundation exam.
Mike
[edit: sorry, after checking my facts we've heard of scores up to 443 and I've seen CEM examples online which indicate it might be possible to score over 500 in a CEM exam.]
Mike
[edit: sorry, after checking my facts we've heard of scores up to 443 and I've seen CEM examples online which indicate it might be possible to score over 500 in a CEM exam.]
Statistically 4 children in 1,000,000 could get a score of 500 - however, I suspect that the marking system is too granular and every year a number of kids achieve a perfect score effectively capping the maximum score somewhere in the 440 - 450 rangemike1880 wrote:Warks is now a CEM exam; CEM exams are often (but not always) standardised around 100/15 SD but not capped at either end, so scores of 143 (or presumably zero) are entirely possible. In Birmingham there have been cases of people scoring more than 500 over the three categories in the KE Foundation exam.
Mike
[edit: sorry, after checking my facts we've heard of scores up to 443 and I've seen CEM examples online which indicate it might be possible to score over 500 in a CEM exam.]
Now the statistics become more complex as we are considering statistical sampling errors - however I can give a very broad rule of thumb here:Charlotte67 wrote:mattsurf, great to have a boffin on board - in the nicest possible way of course!
Some things that I have been wondering about:
1) How large a sample is needed to make these statistics meaningful. eg in an exam cohort of 700, where the children are only compared against those with the same birth month, we are talking about an average of 60 children per month.
2) If each month is standardised seperately is there any advantage to having been born in a less/more popular month?
Help. My head hurts when I think about statistics!
Charlotte
If we have 100 samples, the sampling error will be around 10%
If we have 1000 samples, the sampling error will be around 3%
if we have 10000 samples, the sampling error will be around 1%
The actual sampling error is dependent on the st deviation, however the samples above are a reasonable guidance
I had assumed that NFER consider the whole population who take the 11+; if in one region only 700 take the test then the sample size per birth month is less than 60 - this is far too small a sample to standardise with any degree of accuracy. If we consider every child who takes the NFER test (maybe 25,000 kids - this is a total guess), then the sample is 2000 samples per birth month so the sample is statistically valid >97%