|
The Use of Interrupted Time Series Designs for Program and Policy Evaluation
|
| Presenter(s):
|
| Joseph Stevens,
University of Oregon,
stevensj@uoregon.edu
|
| Keith Zvoch,
University of Oregon,
kzvoch@uoregon.edu
|
| Drew Braun,
Bethel School District,
dbraun@bethel.k12.or.us
|
| Abstract:
The proposed paper describes the use of interrupted time series (ITS) designs for program evaluation. Using multilevel, longitudinal models (HLM), we evaluate whether instructional interventions, student characteristics, and district policies affect either the level or growth of student reading fluency. ITS designs are often recommended in quasi-experimental situations (Shadish, Cook, & Campbell, 2002) as an alternative method for providing control over threats to internal validity. Multilevel, longitudinal modeling provides a flexible and powerful means for analysis of these designs. Another advantage of this analytic approach is that intervention need not occur at the same time for each student. By using individual growth trajectories, the effects of intervention can be modeled at the time of occurrence. The paper will demonstrate the use and application of HLM for the ITS design and provide examples of the estimated relationships between intervention, student characteristics, and district policies with level and growth in reading fluency.
|
|
Multilevel Analysis of Pay Equity: Individual and Organizational Factors Within a Single Organization
|
| Presenter(s):
|
| Blair Stephenson,
Los Alamos National Laboratory,
blairs@lanl.gov
|
| Abstract:
Traditional human capital approaches to understanding pay equity have relied primarily on methods such as OLS regression, where individual factors are emphasized. Multilevel analysis methods more readily accommodate factors related to organizational structure, resulting in more accurate parameter estimates, as well as the potential for more practical and flexible model(s). Decisions regarding higher level(s) may necessitate a tradeoff between statistical and pragmatic concerns. Comparisons to OLS approach, estimates of random group effects, and issues related to the use of multilevel models in pay equity studies are discussed. Note: to ensure confidentiality, as the study dataset is derived from a single organization, some aspects of the underlying dataset and specific findings have been modified. The scope, methodology, and issues encountered during the study are accurately portrayed.
|
|
A Hierarchical Linear Modeling Approach to Analyzing Reading Fluency Growth
|
| Presenter(s):
|
| Steven Guglielmo,
University of Oregon,
sgugliel@uoregon.edu
|
| Joseph Stevens,
University of Oregon,
stevensj@uoregon.edu
|
| Drew Braun,
Bethel School District,
dbraun@bethel.k12.or.us
|
| Abstract:
The proposed paper uses a hierarchical linear modeling (HLM) approach in analyzing patterns of growth in reading fluency over time. HLM is a powerful method in the analysis of change, allowing one to describe growth in terms of two basic parameters: starting values (i.e., intercepts) and rates of change (i.e., slopes). In addition, these models can incorporate multilevel variables—including both student- and institution-level characteristics—as predictors of intercepts and slopes. Using an HLM framework, we examine whether the patterns of reading fluency growth differ as a function of students' gender and ethnicity. The paper will apply various HLM models in examining growth in reading fluency, demonstrate significant relationships between student characteristics and reading growth, and discuss implications for school policy.
|
| | |