We designed the Baby FACES sample to be representative of the population of Early Head Start programs at the national level. Within programs, the families being served by Early Head Start represent the population of parents of newborn and 1-year-old children enrolled in Early Head Start in spring 2009. To achieve the goal of an efficient, representative national sample of sufficient size to detect developmental or programmatically meaningful differences over time and within key subgroups, we use a stratified clustered sample design. We selected all children receiving centerand/or home-based services from a probability sample of Early Head Start programs (including those receiving services through partnership arrangements). Children whose date of birth (or due date, for expectant mothers) fell within the study-defined windows were selected for the Baby FACES sample. We oversampled larger programs to yield more children and families.
The sampling frame we used to select programs was the most recent Head Start Program Information Report (PIR) data, which at the time of sampling covered program year 2006-2007. All Head Start and Early Head Start programs must submit PIR data annually. For this reason, the PIR is a reliable source of information about the programs and the type of families they serve (for example, PIR data include the number of children served and the demographic characteristics of families in each program). The sampling unit was the PIR reporting level—that is, the grantee or delegate agency (or “program”).