So, no guarantees that this is the explanation, but I'll throw a few thoughts and general guidelines out there.
First, a couple of little housekeeping items: some of your columns are reversed; the column with the 100/90s is actually the RPI, and the previous column, that puzzles you, is the standard score. The SS is a deviation measure, comparing your performance to an age-normative group in a rank order manner (related to percentile). The RPI compares how challenging tasks might be expected to be. It has a very low ceiling, as the second 90 represents the level at which the median peer would be expected to be 90% successful. On that task, your student would be predicted to be the first number % successful (in this case, 98 to 100%).
Second, Broad Reading is derived from LWI, RF, and PC. Brief Reading omits RF. WA only comes into the Basic Reading Skills cluster, which you have not reported.
Now we get to the odd part of things. Deviation scores (SS, in this case) represent both the distance from the norm, and the rank order in which a test performance occurs in the norm population. When you start combining performances (i.e., subtests) to form composites (i.e., cluster scores), sometimes they don't come out the way you would expect. I'll name two factors that often play into this:
1)The likelihood of any given person having an unusually high score in one isolated area (a splinter skill, as it were), is actually much higher than one would think from a common sense standpoint. What's really unusual is scoring in that same high range across the board. So the cluster/composite score derived from two 150s might be expected to be higher than that derived from a 160 and a 140, as it was a rarer occurrence in the norm population, even though the mean score of both is 150. Not all tests derive their composite scores this way, and, to be honest, I can't remember if the WJ is one of them, but it is possible that this is the case.
2) Out at the extremes of the norm group, there are fractional individuals representing these standard scores in the actual norming population, so test developers have to use statistical smoothing methods to estimate standard scores. (If you have 2000 people in your norm group--which is considered a pretty good size--and your norms are divided into, let's say, three groups per year for school-age children, this results in roughly 50-60 children per group, which means that one child is representing the entire top 2% of the population. You can see how imprecise this is for the extremes of the bell curve.) Now, the statistical estimation methods are better than one would expect, but still, once you get out beyond three or four standard deviations from the mean (145+), it's not very well connected to the actual standardization sample.
For your child, there is a pretty wide range among the subtests feeding into the cluster scores, so it is possible that either or both of these factors were involved.
Last edited by aeh; 05/13/14 10:25 AM. Reason: started this on an iphone.
