The National Center for Education Statistics released July 14, 2006, a report Comparing Private Schools and Public Schools Using Hierarchical Linear Modeling.
In grades 4 and 8, using unadjusted mean scores, students in private schools scored significantly higher than students in public schools for both reading and mathematics.
But when school means were adjusted in the HLM analysis, the average for public schools was significantly higher than the average for private schools for grade 4 mathematics and not significantly different for reading.
At grade 8, the average for private schools was significantly higher than the average for public schools in reading but not significantly different for mathematics.
Comparisons were also carried out between types of sectarian schools.
In grade 4, Catholic and Lutheran schools were compared separately to public schools. For both reading and mathematics, the results were similar to those based on all private schools.
In grade 8, Catholic, Lutheran, and Conservative Christian schools were each compared to public schools.
For Catholic and Lutheran schools for both reading and mathematics, the results were again similar to those based on all private schools.
For Conservative Christian schools, the average adjusted school mean in reading was not significantly different from that of public schools.
In mathematics, the average adjusted school mean for Conservative Christian schools was significantly lower than that of public schools.
This summary appears consistent with common sense application statistical principles. It takes great effort to override the power of large numbers when calculating mean scores for aggregates.
I haven’t read these reports, but in skimming portions, I see no reference to methods of instruction and use of learning devices such as a Tablet PC that might contribute to different student scores.
Again, using statistical principles, it seems reasonable that electronic learning devices would likely change student conditions sufficiently to alter mean scores for that subset of all students.
Hmmm. Well, at least that’s a hypothesis worth testing. Please point me to studies that have tested the null of that hypothesis.