The work samples of students (n=45) were analyzed at Time 1 (pre-test), Time 2 (embedded assessment), and Time 3 (final assessment), using linear growth modeling (Singer, 1998; Bryk and Raudenbush, 1992; Willett, Ayoub, and Robinson, 1991). Growth modeling, in which data are collected longitudinally, or in multiple waves, provides a vastly superior way of looking at the gains students make, and allows you to examine the pattern of growth over time. It increases the power of the sample because there are multiple data points for each student (Willett, Ayoub, and Robinson, 1991). This modeling process informs our understanding of patterns of academic growth of individuals (within-child growth) and of the student population as a whole. It also provides an opportunity to model predictors of growth and its variance as a function of a variety of student characteristics. Using linear growth modeling, we were able to determine student growth on each measure over time, testing for any variance based on the amount of support a student needed (i.e., IEP, at risk, or typical).

The sample also included the work of three students with severe cognitive delays. This student work was not included in the statistical analysis; rather, at the request of the school faculty, these students were assessed alternately, and the result of this assessment is reported separately.

To inform the statistical analysis and reinforce the quantitative data with qualitative data, we reviewed teachers' perceptions of the impact on students through analyzing transcriptions of the final action reflection meetings where teachers discussed what science concepts students had learned. Teachers were identified and coded, using the qualitative software NUDIST (Sage Publications Software). In this way, we were able to identify and isolate teachers' comments about student growth. We also analyzed teachers' perceptions of what students learned from the final evaluations. Finally, we reviewed the progress notes and IEPs of students with disabilities to determine if any modifications had been added to their programs or if science goals were noted.

To determine the impact of the action reflection process on teachers, special educators, and curriculum specialists, we analyzed the transcription data qualitatively. Broad coding categories were developed from the research questions. Several members of the research team independently coded selected transcripts of action reflection meetings to develop specific codes within the broader categories. They then compared codes and resolved any discrepancies. We created a "code map," using the NUDIST coding structure, to include these broad categories and specific codes. Using the code map, two researchers independently coded each data document, then reviewed the document together and came to consensus. Using codes that enabled us to identify statements by role and type, we identified themes in how teachers and special educators described the impact of the process. We reviewed the data and the emerging categories with the research team and compared evidence from the action reflection transcriptions to the evaluation data from the focus group. We found that teachers' and specialists' comments about the Action Reflection Process fell into four general categories -- changes in their roles, changes in practice, changes in assessment, and impact on students -- and we based our conclusions on this data. To confirm that teachers actually implemented the suggestions for supporting students with disabilities, we reviewed the classroom checklists and student work to note the strategies that teachers implemented most frequently.

To determine the impact of the Action Reflection Process on the district, we analyzed the documented evidence of dissemination of the process. Through numerous conversations with Lin Tucker, other science staff development teachers, and principals, we learned of the use and dissemination of the process throughout the district. We also continued to maintain discussions with all three principals after the funding period, and we documented their reports of efforts to sustain the process within their schools, without the presence of the researchers.