~ ReNcH ~ said:
Are raw marks evenly distributed between the cut-offs?
Yes. A simple linear transformation is used to interpolate marks between cut-offs.
I'll demonstrate it using your example. You know that 60 is aligned to 80, and that 80 is aligned to 90. We therefore have two pairs of marks corresponding to the endpoints: (60, 80) and (80, 90). The raw mark is the x-coordinate, and the aligned mark is the y-coordinate. The straight line through these two points provides the relationship between all raw and aligned marks that fall within that band.
There is a separate straight line for each band, because each band has its own endpoints.
~ ReNcH ~ said:
Also, could several different raw marks attain the same aligned mark? For instance, could both a raw mark of 117/120 and 118/120 attain an aligned mark of 99? Or would the 117/120 automatically attain an aligned mark <99?
It depends on the precision used. The Board transforms raw marks directly to aligned marks, and hence the aligned mark corresponding to a raw mark of 117/120 is technically less than the aligned mark corresponding to a raw mark of 118/120. But as jm123 said, because aligned marks are rounded to the nearest integer, it appears as though the two raw marks correspond to the same aligned mark.
~ ReNcH ~ said:
Is there really much point in aligning marks to an accuracy of 0.1? I put some marks into SAM, and 0.1-0.5 barely makes a difference to the overall UAI estimate.
So in other words, 117/120 could be 98.6, while 118/120 could be 98.9? Both would be reported as 99.
This is largely dependent on the magnitude of the aligned mark - an additional 0.1 will make a much larger contribution when the aligned mark is below 50 than when the aligned mark is above 90, for example.
In practice, little is achieved by using the extra precision. I only use it when attempting to determine a particular percentile or scaled mark exactly.
