Martine van der Pluijm

104 Chapter 4 counted the number of parents that entered the classroom and the number of parents that participated in the parent-child activities. Next, we coded teacher behavior on the four steps of the program (Appendix A). We used a 5-point scale to indicate the frequencies of the observed behavior or the quality of the behavior ranging from 1 (none of the behavior observed) to 5 (continuous behavior observed). At the pretest and posttest, the first author conducted three of the 13 activities with one of these assistants and four with the other assistant. The first author and the assistants coded the same activities independently using the observation scheme. Intercoder agreement was calculated as a percentage of agreement for each of the pairs of coders. Percentages of at least 80% agreement at pretest and posttest were considered adequate. Analyses We transcribed the 14 interviews collected at the posttest in three parts to answer research questions 1 and 3. First, we coded teachers’adherence using amatrix to sort utterances that were related to the seven steps. We distinguished teachers’ adherence on four levels: no adherence, some adherence, convincing adherence, strong adherence. We classified teachers’ adherence as convincing if we found at least one aspect of the step showing teachers’ integration of that aspect in practice, and strong if we found two or more aspects of a specific step. We summarized all the codes in a table that shows teachers’ perceptions of their adherence on three levels. Second, we coded teachers’ utterances using two matrixes with categories that referred to teachers’ intrinsic motivation (perceived autonomy, relatedness, and competence), their perceived improvements of parental roles and children’s development, and their intention to continue using the AHL program. Again, we summarized all the codes for each of the teachers. This resulted in a second table that shows teachers’ perceived goals. The interrater reliability was calculated using Cohen’s Kappa. The average score was at least Kappa = .76, which can be considered as substantial agreement. Two final tables were developed with a summary of codes for each teacher. For the questionnaires and observations, we analyzed improvements of teacher adherence to the steps of the program from pretest to posttest. We used repeated measures ANOVA to analyze progress in teachers´ insight in parental knowledge and skills and the HLE of each of the parents in their classroom (Step 1) and their self-reported ability for building reciprocal relationships with them (Step 3) based on the class inventory list. We used education levels of the parents in the classroom of each teacher as a factor to explain changes from pretest to posttest. We checked for equality of error variances and covariance matrices in these analyses. No violations of the assumptions for this analysis were found (Levene’s and Box’s test). We used t-tests to analyze teachers’ progress on the remaining steps of the AHL program (Step 2, 4, 5, 6, 7). Finally, we calculated Cohen’s d effect sizes for our comparisons of pretest and posttest results.