Calculation of Scaling Using A3 data (1 Viewer)

[Captain pi]

Is it possible to calculate one's scaled marks from the data supplied by UAC @ http://www.uac.edu.au/pubs/pdf/2004-Table-A3.pdf?

The way I would calculate it is, taking this hypothetical example.

Subject 1 U has an HSC mean of 25.0 and a SD of 5.0. Its scaled mean is 30.0 and its scaled SD is 10.0.

If someone got an HSC mark of 30.0 (SD = +1.0), is their scaled mark 40.0? This seems logical to me, (I'm no statistician), but I get marks which are slightly different to the ones released when I compute the 99 percentiles, 90 percentiles, etc.

Thank you.

 

[Lazarus]

You're attempting to apply a linear transformation to determine your scaled mark from your aligned HSC mark.

There are two problems with this:

1) Scaled marks are determined from distributions of raw HSC marks, not aligned HSC marks. You would therefore need to know your raw HSC marks and the raw mean and standard deviation for each course.

2) A non-linear transformation is used instead of a linear one. Even if you had the raw data, you would need two additional parameters (let's call them "a" and "b") that have been calculated for each course. You can't calculate them yourself without having the entire distribution of raw marks.

It is possible to approximate scaled marks using Table A3 via various methods of interpolation.

 

[Captain pi]

Okay, so my method is flawed; I thought so. Thank you.

(Where do you get your data for SAM, Lazarus?)

 

[phatic]

Don't you just convert the HSC marks into scaled marks (which are shown on that table), then add up all the scaled marks, and convert that number into a UAI using table A8? Of course, approximating the values in between the percentile readings. I found this method fairly accurate when applied to my own UAI... I got within one or two points, I mean.

 

[Captain pi]

In theory? Yes, I believe that is what you do. However, because not all the marks are given, and because they are not linearly transformed (didn't you know that?! Jeeez.), it is unlikely that you would be able to determine your aggregate from this table (unless all your HSC marks happened to be in the table; that is, you were one of the given percentiles for each course.)

 

[phatic]

Yeah, that's what I thought. Not exactly accurate, but close enough for me. I just approximate the values in between the percentile figures as if they were linear. I figure I can't be that far off, not when we're talking about one or two marks.