PIPS:Why don't you use maths to predict maths, and reading to predict reading?
Posted by Chris Jellis on 22 July 2011 10:45 AM


We use a (weighted) average of maths and reading as a predictor because this predictor correlates better with future attainment in maths (and predicts more accurately) than maths alone does, and it correlates better with future attainment in reading than reading alone does. We choose the weighted average based on empirical evidence – it is in general the better overall predictor. If a child does better at reading than maths, they tend to still do better in reading next time. However, using reading to predict reading and maths to predict maths gives us exactly the same problem. If two children have the same high reading score, and we predict using reading only, we predict they will both get the same reading score as each other next time. However on average a child with a similarly high maths score will do better in reading than a child with a comparatively low maths score. Similarly, if two children have the same maths score we may expect them to end up with the same maths score, however if one child has a comparatively much higher reading score they will make more progress in maths. So, if a child has a higher reading score than maths score, using maths to predict maths may increase the maths valueadded, but using reading to predict reading would correspondingly lower the reading valueadded. Using the weighted average of maths and reading scores is therefore the best overall predictor.  
