Analysis Methods

Impacts on classrooms and instructional practices were analyzed using two-level hierarchical models where classrooms (level 1) were nested within randomization blocks (level 2). Treatment impacts were estimated in models that controlled for year 2003 baseline measures of teacher behavior and classroom environment, dominant language of teacher and size of center. Impacts on children’s language and literacy skills, as measured by the 2005 TOPEL, were analyzed in three-level hierarchical linear models in which children (level 1) were nested in classrooms (level 2) and classrooms were nested in randomization blocks (level 3). Treatment impacts were estimated in models that controlled for child’s age, sex and language spoken at home, for classroom-level mean LAP-D Cognitive Total scores measured in Fall 2004, dominant language of teacher and size of center. More detailed information on the specification of models is provided in Attachment A.

The impact estimates presented in this report represent: first, the effects of the three treatments on teacher behavior and the classroom environment after six and eighteen months of participation in the study; and secondly, the effects on children of between two and ten months of exposure to the curricula.