uaiseeker question (1 Viewer)

[spin spin sugar]

i just KNOW you're going to tell me "Well, you can't say, it depends on the group of students.." etc etc but im going to ask anyway! so try and go out on a limb and give me an educated guess!!

basically, i'm curious about the 2001, 2002, and 2003 predictions.
for the sake of my question, these are my predictions:

2001 - 97.50
2002 - 96.85
2003 - 97.25

i'm curious as to what the 2003 prediction is based on. it's higher than 2003, yet the drop in uais from 2001 to 2002 would suggest that students are getting BETTER -- and then wouldn't uai's be lower overall...? because it is a 'rank'? if that's true.. then why is the 2003 estimate higher than the 2002? are the 2003 estimates based in any way on our year's year 10 school certificate results?

also.. which estimate would be more accurate?? i tend to use the 2002 predictions the most, i always just assumed it would be more accurate. is that a fair assumption to make? i KNOW it all depends on this year's students' results... but basing it ALL on assumption (which i know is a silly and risky thing to do)

a) what year prediction is likely to have the most accurate estimate?
b) why is the 2003 prediction higher than 2002??

 

[Jeo]

a) 2003, since they claim its accurate. Anyway, its to give you an estimate, not your exact UAI.
b) Probably because im class of 03.

 

[Lazarus]

a) Well, you can't say, it depends on the group of students... I would be inclined to suggest the 2003 projection, but in reality it's simply not possible to obtain the degree of accuracy you're after. There's only a difference of 0.65 between your highest and your lowest estimate - the uncertainty increases with the precision.

b) The 2003 prediction utilises results from both 2001 and 2002 in an attempt to establish a trend. Obviously, it's rather difficult with only two samples, but it's the best that can be done at the moment.

 

[spin spin sugar]

thanks laz. thats helpful

another question - would you recommend using the percentiles estimate over the marks one?

 

[Lazarus]

Originally posted by spin spin sugar
another question - would you recommend using the percentiles estimate over the marks one?

Yes. Try comparing predictions over different years using the same percentiles to predictions over different years using the same marks. You'll see why.