ti To Aggregate Is to Aggravate *

Aggregation of data is, by definition; an obscuring of details for the sake of achieving a summary. It is, therefore, potentially harmful to accuracy. Attention should be given to the aggregation that scores undergo prior to statistical tests. Two familiar research designs where this is important are a) in two groups/one measure cases, and b) in two groups/pre-post measurement cases. Another problem for researchers develops when incomplete and missing data are encountered for identification codes, as well as for score values. Unless each data record contains all identifying codes, it will be excluded from one or more aggregates and results at different levels will chance depending on the data values for those records which are missing identification codes. Observations recorded on the Individual Cognitive Demand Schedule can be examined as an illustration of the problems with aggregation. Although aggregating the data is generally beneficial in this case, it leaves out considerable evidence about the classrooms. Thus, the value of these data is greatly enhanced by leaving the observations unaggregated. (PB