Median Split of Sample

Advantages of a Median Split

The advantage of splitting a sampling into two equal groups is the ability to employ the variable within various designs as an independent variable. A lot of statistical analyses involving commonly used statistics (ANOVA, t-test) are more convenient and subject to ease of interpretation if the continuous variable becomes transformed into a dichotomous (or categorical) variable using the median score to divide the sample. A variable can be divided into three, four, or any number of levels based on some type of score chosen. The ability to interpret the outcome of a statistical test is much easier, in many circumstances, if the process of data reduction provides for a reduced set of options and interpretations.

Often, the variable used to divide the sample reflects the measurement of some individual difference characteristic. For example, the PRCA measures how persons vary based on level of apprehension, and this difference between people may be used to predict some outcome or dependent variable, like willingness to run for elected office. In a complex design that desires to consider biological gender (male and female) as well as level of education (college graduate or not), the combination of variables (in this case three—PRCA, biological gender, education) generates a simple ANOVA design. The particular ANOVA design specified in this example would be a 2 (gender) × 2 (education) × 2 (PRCA) with a total of eight possible combinations or cells for analysis. The design permits the examination of some types of nonlinearity or interaction analysis among the independent variables (gender, education, PRCA). [Page 975]The reliance or preference for an ANOVA analysis is easier if the continuous variable is reclassified and represented as a categorical variable based on a split using the median.

The advantages for the use of the median split stem from the ease of using familiar and relatively well-defined statistical procedures commonly taught. The application of the technique essentially makes the task of statistical analysis easier and interpretation of any results more obvious and simple. The technique requires little background or difficulty and most statistical packages readily permit use of the technique. The redefining of the variable from continuous to dichotomous should not provide essential problems in terms of the actual mechanical operation.

...