Effect Sizes of Behavioral Interventions for Turkish Students with Autism

Sucuoglu B., Demir S.

EGITIM ARASTIRMALARI-EURASIAN JOURNAL OF EDUCATIONAL RESEARCH, vol.12, no.48, pp.21-38, 2012 (SSCI) identifier identifier


Problem statement: Two of the techniques used in the meta-analysis of single-case research are calculating the percentage of nonoverlapping data (PND) and the percentage of data points that exceed the median of the baseline phase (PEM). The PND is a nonparametric approach in which the researcher calculates the percentage of data points in the treatment phase that exceed the highest point of the baseline phase's distribution (or the lowest point if the targeted behavior is a problem behavior to be reduced). Even though PND can be calculated easily, it has a disadvantage if there is even one data point in the baseline phase over the top or below the bottom level, the PND score for that treatment phase is calculated as zero. Calculating the PEM was introduced in order to overcome this disadvantage. In this approach the effect size is calculated by finding the median of the baseline phase and drawing a horizontal line from this point through the intervention phase and then determining the percentage of data points over this line (or below this line for the problem behaviors).