Page 1 of 9

exam strategy (+ bit of prediction fun)

Posted: Fri Sep 18, 2015 8:26 am
by Petitpois
My DD has provided me a break down of each section in the exam. The number of questions in that section. How many she answered. I obviously know when she was born and have a reliable set of stats on her success rate in a variety of timed tests in the run up to last Saturday, i.e. the average % she was getting on each type of paper. I won't say what these were either.

My predicted score based on missing questions in all sections (apart from maths) + and known averages from her bond papers is .......................................

194

So I think she we will be about 15-20 below qualifying score. Potentially disappointing but a harsh lesson that your up against the best in Birmingham, and you just don't have the comfort zone to be throwing away potential marks by not answering questions.

It will be really interesting to see if there is any merit in the comments about missing 28 questions and still scoring highly enough to get a place.

Before every one starts getting all scientific - just treat this as a bit of fun. I am aware of all the variables and margins of uncertainty that exist. I really struggled to understand whether we were making good enough progress over the last 18 months. I hope this might prove useful to others starting the journey. BTW also give you her levels from end of Year.

We did try and ensure that DD answered all questions, but for whatever reason it did not happen. DD did not listen and just make guesses with the final questions in each section, when the 1 minute warning was given. Many many children will have done that (DD's best friend, did).

Question to forum - is the optimal exam strategy to work through questions as quickly as possible and then just guess remaining ones when time is running out????

I suppose if you're considering it, you are trading 1 further minute at your normal methodical pace, (let's say 3 questions for Joe average), where you are able to make your best shot Vs for all remaining questions and having a 1 in 4 or 5 chance of getting the correct answer.

It is all clutching at straws now though, bottom line is that she should have been able to answer questions accurately and at speed, for all sections, to be in with a chance against the best of 5,300 competitors.

So in conclusion I will be raising a cheer for all those that pass, regardless of exam strategy they employed. They will have deserved their success!

I will come back on Oct 16 and post a reply here if moderator allows, detailing each section and final outcomes, as well as average bonds scores. Plus bond placements test scores and levels results from Sep 14 and Mar 15 for DD.

Re: exam strategy (+ bit of prediction fun)

Posted: Fri Sep 18, 2015 8:44 am
by MSD
How did you get down to that number 194 :) 200 would be an average score and I had a formula somewhere that worked out how many children will achieve above a certain standardised score based on numbers sitting.

Even if you miss 50 questions, but do well on others there is good probability to achieve above 194.

Let me dig that formula out!

Re: exam strategy (+ bit of prediction fun)

Posted: Fri Sep 18, 2015 9:17 am
by MSD
Here you go:

This is one of my posts from many years back.

"I was provided with the following formula by Okanagan to approximately work out the percentage of candidates who scored above a certain standardised score and here it is:

Let's consider a standardised score of 339 - Assuming it was made up of three 113 scores that would give you a z of (113-100)/15 = 0.8666 for each subject area
0.8666 equates to approximately 20% scoring higher for that subject area (Try this
http://www.doe.virginia.gov/testing/tes ... _table.pdf" onclick="window.open(this.href);return false; z-table)

so 20% of the total candidates would be ranked higher.

But as we don't know the degree of correlation between the three subject areas - working out the three separate ranks and averaging them would probably give as good an estimate of overall rank as you are going to get.

Hope this helps!"


Now let's look at your number 194. As Brum is now divided in two subject areas, so in this specific instance just as an example if we divide it by 2, you will get a score of 97 in each subject area.

(97-100)/15 = -0.2

From the attached Z table you would be 42 percentile which would mean 58% of children scoring higher than a score of 194.

So it will be approximately around that number scoring higher. That will be 3190 children out of 5500 scoring higher than her based on your predicted number.

Re: exam strategy (+ bit of prediction fun)

Posted: Fri Sep 18, 2015 9:18 am
by Petitpois
Very guessy, but sorted of informed guess. Approach was

Take total number of questions in exam - assume 1 mark for each question (Does anybody know for what the marking schema is - do CEM weight some questions or just award higher marks for some question types?)

Derived a notional maximum marks for my DD.

Multiplied notional maximum * average bond scores for each question type e.g. maths = Guessed RAW score (ie a % of notional max)

Took neighbours standardised score. And his DC notional maximum marks (assumed last years and this years were broadly the same) - Applied average bond score to guess his probable RAW. Compared the two and came up with 194. Making a further assumption about age effect on standardised score.

Bit of fun, and you can see people laughing heartily out there, at my Voodoo cum Quackery statistical attempt to guess. I know, I know just wait till they come out

Re: exam strategy (+ bit of prediction fun)

Posted: Fri Sep 18, 2015 9:23 am
by Petitpois
MSD there were two papers covering multiple subjects, is that what you mean't??

Re: exam strategy (+ bit of prediction fun)

Posted: Fri Sep 18, 2015 9:25 am
by MSD
Yes, so just an example I have divide the total score OP provided by 2 as Brum now standardise two ways - 50% for English and 50% for Maths and NVR.

Re: exam strategy (+ bit of prediction fun)

Posted: Fri Sep 18, 2015 9:40 am
by Bob1892
I've managed to concoct a formula thanks to posters like vza, Okanagan and KenR to derive a final sd score. I'll let you know in 2 weeks time when we get the Walsall results whether it was accurate enough :?
It certainly worked for me 3 years ago when my dd took the brum and warks cem tests- yes, the same dd who missed out 28 questions and still scored 240+. I was within 3 sd points for both tests.
It helps if your dc gives you a breakdown of the test.
My ds is capable of knowing the number of questions in each section, how many he thinks he might have answered correctly (on the whole he's been fairly accurate with this over the last few months). He rated each section out of 10 in terms of difficulty.
However he hasn't given me much info about test content though :roll:

Re: exam strategy (+ bit of prediction fun)

Posted: Fri Sep 18, 2015 9:54 am
by Petitpois
Thanks Bob, all good reasons to continue to hope.

Re: exam strategy (+ bit of prediction fun)

Posted: Fri Sep 18, 2015 10:13 am
by MSD
Bob1892 wrote:I've managed to concoct a formula thanks to posters like vza, Okanagan and KenR to derive a final sd score. I'll let you know in 2 weeks time when we get the Walsall results whether it was accurate enough
Bob - You really have to be a magician to work out the actual SD score for your DD :)

I can see you have a good understanding of how she did in the exam and what she might have scored in terms of raw mark. But to work out the actual SD you need to have a fairly good idea of how the other children have performed as compared to her and how many have done better or worse then her.

Yes, if you know the actual SD score and the total children sitting the test you can approximately work out the percentile and rank. But accurately predicting SD score is entirely another thing. Well established tutors can possibly, based on years of experience, roughly predict what your child might score as compared to the other children they have been tutoring or have tutored in the past. And this also depends on how your child performed on that particular day!

Re: exam strategy (+ bit of prediction fun)

Posted: Fri Sep 18, 2015 10:54 am
by nervousmom
PP, as you have managed to get more info from your DC than I seem to have got from mine! Do to have a rough idea of how many questions there were in total, or the base from where you worked?

I've been working on a base of 300 :? DD reckons she missed 2 or 3 and guessed about 20. So 300 - 23 puts her down to 277. We have assumed that she got 73.4% of the questions right that she answered(this was the average score of the 5 mock exams she did) and have assumed that of the 20 guessed, 25% would be right, giving her a raw score of 208! ( however if out of 310 she would get 215, out of 320 she would get 222, out of 330 would be 229 and so on)

There is a lot of room for error with my theory :shock:

If anyone can expand on this it would be useful :)