2010 PSLE Discussion

[atutor2001]

atutor2001:

My analysis to estimate the average and stardard deviation based on feedback we got here :


T-score 282 (assuming that the raw score is 99 for every subject)
T-score 263 with 4A* (assuming that the raw score is 91 for every subject)

Based on the above assumptions, the \"average\" mean per subject is about 64 and the \"average\" standard deviation i.e. spread is about 17 marks.

Very surprised that the sd (spread) is so big.

There is another case where 3A* & 1A gets 249. Based on my above estimates for mean and SD, the T-score will be about 254. Therefore, the above estimates are quite off.

I have adjusted the assumption for the average raw score from 91 to 92 for the T-score of 263. That gives the following new estimates of the mean and SD :

Mean is about 68.8
Sd is about 14.7

With this mean and SD, the 3A* & 1A will be about 249.5.
Therefore, I think for future estimation of T-score, we can use

Mean = 69 and SD = 15

[atutor2001]

autolycus:

All subjects at PSLE are accorded 'equal weighting' in the sense that the aggregate score is the sum of all the t-scores without any extra 'bias' or 'weighting' assigned to any subject.


For students with no MTL (exempted), the aggregate is computed by multiplying by 4/3 so that the three remaining subjects will now have the 'weight' of 4 subjects.

Hi autolycus

You are right. I used my new estimates for the means and SD and the T-score for the worst case of 1A* & 2A comes out to be 230.22 based on the average of the 3 subjects to get aggregate of 4 subjects.

[mum-to-2]

my ds got into express with a score of 200. he wont be able to get into his affiliated secondary school since the school’s COP is 225. does this mean i should not even list it as his first choice?


schools in our list include:
pierce cop 208
monfort cop 206
holy innocent cop 219
guangyang cop 203
chong boon cop 190
bartley cop 193
ankss cop 227
mayflower cop 222
beatty cop 217

appreciate your advice thanks!

[JonC]

@atutor


Here are few cases I know,

3A. 1B 230 and also 220

3A, 1A* 243 and also 230

Does this help you to fine tune further?

But not every subject will have the same Mean and SD right?

[verykiasu2010]

atutor2001:

atutor2001:

My analysis to estimate the average and stardard deviation based on feedback we got here :


T-score 282 (assuming that the raw score is 99 for every subject)
T-score 263 with 4A* (assuming that the raw score is 91 for every subject)

Based on the above assumptions, the \"average\" mean per subject is about 64 and the \"average\" standard deviation i.e. spread is about 17 marks.

Very surprised that the sd (spread) is so big.

There is another case where 3A* & 1A gets 249. Based on my above estimates for mean and SD, the T-score will be about 254. Therefore, the above estimates are quite off.

I have adjusted the assumption for the average raw score from 91 to 92 for the T-score of 263. That gives the following new estimates of the mean and SD :

Mean is about 68.8
Sd is about 14.7

With this mean and SD, the 3A* & 1A will be about 249.5.
Therefore, I think for future estimation of T-score, we can use

Mean = 69 and SD = 15

don't think this is reasonable

every subject's mean and their SD is different, and it fluctuates from year to year

this year's mean for all subjects are definitely higher than last year as the papers were easy, and the SD must have been narrowed too.

giving you as examples below :

NYPS 2010 Prelim including GEP of about 100, total cohort size 472 :

Mean for EL is 79.97 and SD is 8.99

Mean for CL is 73.47 and SD is 11.53

Mean for MA is 65.80 and SD is 17.93

Mean for SC is 68.69 and SD is 13.45

all rounded to nearest 2 decimal; at least 82% got A / A* for all subjects

NYPS is one of largest Pri 6 cohort and its numbers should be quite representaive

there is also a student with 3A* + 1A = 262, compared with the 249 quoted above

[sautille]

verykiasu2010:

there is also a student with 3A* + 1A = 262, compared with the 249 quoted above

Just to add to the confusion...my son's classmate has 3A*+1A and a T-score of 269 (without HMT).

[verykiasu2010]

sautille:

verykiasu2010:

there is also a student with 3A* + 1A = 262, compared with the 249 quoted above

Just to add to the confusion...my son's classmate has 3A*+1A and a T-score of 269 (without HMT).

no confusion, 262 is close to 269 compared with 249 in the prediction....that 1A of 269 is probably a very very high A, missing the * by the skin of his teeth....

it just goes to show that there is no uniform mean nor SD, every subject every year is different as the individuals sitting of the papers are different

[atutor2001]

JonC:

@atutor


Here are few cases I know,

3A. 1B 230 and also 220

3A, 1A* 243 and also 230

Does this help you to fine tune further?

But not every subject will have the same Mean and SD right?

Thank you JonC for the additional information. The case of 3A, 1A* with 230 helps. (It is consistent with the estimates - the worst case will give a T-score of 227 so 230 is still not the worst case.)

I have been trying to establish a good estimate for the mean and the SD because such info is not available. Through results of those \"extreme\" cases, e.g. all A* but yet relatively low T-score, we can get quite good estimates for the overall mean and SD.

There is no breakdown to allow the estimation for individual subjects. However, from the overall mean of about 69, we can roughly distribute the mean for individual subjects as follows :

Math : 75 (usually the highest)
Chinese : 69 (paper has become much easier over the years)
Science : 65 (many cannot cope with Section B)
English : 67 (result of 4x69 - 75 - 69 -65)

For the Std Deviation, I have been trying using different values and find that the 14<SD<17 will give results consistent with the T-score so far.

Hope it can serve as a rough guide for future parents to work out the T-score.

Most important finding is : \"The passing mark to get into express in Sec Sch, is 70 and not 50 (the traditional passing mark).

[atutor2001]

verykiasu2010:

.....

NYPS is one of largest Pri 6 cohort and its numbers should be quite representaive

there is also a student with 3A* + 1A = 262, compared with the 249 quoted above

Have to disagree with you on NYPS being \"representative\". Any school with GEP cannot be \"representative\".

The range of possible total raw score for 3A*+1A is 348 to 390.

It is normal for 3A* + 1A to get 262 T-score but is of no help for estimation.

However 3A* + 1A with 249 T-score is representative of an extreme case i.e. possible worst case scenario which gives us an indication that total raw score could be 340 (worst case). By collecting all these cases that can give reliable indication of the raw score, we can estimate the overall mean and SD.

[verykiasu2010]

atutor2001:

verykiasu2010:

.....

NYPS is one of largest Pri 6 cohort and its numbers should be quite representaive

there is also a student with 3A* + 1A = 262, compared with the 249 quoted above

Have to disagree with you on NYPS being \"representative\". Any school with GEP cannot be \"representative\".

The range of possible total raw score for 3A*+1A is 348 to 390.

It is normal for 3A* + 1A to get 262 T-score and is of no help for estimation. However 3A* + 1A with 249 T-score is representative of an extreme case i.e. possible worst case scenario which gives us an indication that total raw score could be 340 (worst case). By collecting all these cases that can give reliable indication of the raw score, we can estimate the overall mean and SD.

every school's raw score is of no help in the estimate because their standards are different from school to school. A child from school A with raw score of 380 cannot be compared with another child of the same raw score from school B.

And nobody actually know the raw scores in the actual PSLE exam, but it is a known fact that every subject's mean and SD are different

please don't be upset if I say the estimate is futile.

it will be more instructive if the schools provide its own t-score for every student every year and bench mark it to actual performance at PSLE and see where they stand. that will be much more able to predict the PSLE actual scores

for example, a few schools know for sure the range of t-score improvement from prelim to actual, every year, based on past experience, how they prepare the students, what standard of questions they practice, what prelim standard they set, etc...