External validity and the choice of statistical method: research in reading

Abstract

Research conclusions which are derived from the application of inappropriate statistical methodologies can lack accuracy and by doing so can lead to confusion and wasted effort. Heterogeneity, which tends to be greatest when special needs pupils act as subjects, is an anathema to the simpler statistical models. A data set is analysed via a number of commonly used statistical methods which take as their foundation the classical notion of independent and dependent variables. An example of an analysis using a technique which assigns subjects to fairly homogenous sub groups is also presented. It is argued that the latter technique better accounts for heterogeneity in a data set and ultimately provides increased accuracy in the explanation of relationships held within that data set. This argument is applied to research in reading.