Robin van Rijthoven

116 Chapter 5 In order to rehearse the above-mentioned spelling and reading knowledge children had to do home exercises for reading and spelling. Parents were asked to train four times a week during 15 minutes with prescribed exercises. All parents have confirmed that this has been complied with. Parents reflected on the home exercises in a day-today logbook. When a child reached an accuracy of 80% during practice (read or write 80% of the words correctly) and improved significantly in their fluency (more fluent compared to the first time words were read) the clinician moved on to the next topic of intervention. This formative testing was sustained throughout the entire intervention. Therefore, variation in the length of the program is present. Analytic approach By means of regression analyses intercept and slope of the five trials of the 15 words test were calculated for each participant to quantify verbal learning. To check whether groups differed on verbal learning and consolidation, two Repeated Measures ANOVA’s were conducted. To answer the research question and maintain enough statistical power, in total twelve separate hierarchical regressions were conducted; four for each outcome measure (i.e., word decoding, pseudoword decoding, and spelling). The first and second set of hierarchical regressions tested the effect of verbal learning on the outcomemeasure before and after the intervention, the third and fourth set of hierarchical regressions tested the effect of verbal consolidation on the outcome measures before and after the intervention. In all cases, relations were tested without control variables first and later control variables were added separately. Based on the fact that verbal learning and verbal consolidation are separate processes (Helmstaedter et al., 1997) we assumed that analyzing the effects of these factors separately would be a good way tomaintain enough power as in the case of the response to intervention only 52 participants were present. In line with Harris (1985), who reported that analysis should include an absoluteminimumof tenparticipants per predictor, thismeant we could includefive predictors atmost. Inorder to express the response to intervention, the individual mean change between pre- and posttest (with percentile scores) was calculated by subtracting pre- from posttest scores. Following Gollwitzer and colleagues (2014), change scores are reliable if two requirements are met. First, standard deviation must differ between measurement occasions. Second, there needs to be a non-zero variation in observed difference scores in order to define the reliability. In order to rule out the effects of variation in length of the intervention, the individual change score was divided by the number of sessions the intervention for each individual lasted. Working with change scores implied that the differences in individual pretest scores were not taken into account. Pretest scores were included as a control variable to control for variation.