NAEP Technical DocumentationVariables that Define Groups in NAEP

The variables used to define groups for a given assessment scale (or group of scales) include a broad spectrum of contextual variables and composites of such variables. All standard reporting variables are used in the population-structure models. Results for any variables not used in the models may be biased.

The initial step in the construction of variables used in the scale score distribution models involves forming student-based vectors of response data from answers to student survey questionnaires; teacher questionnaires; school questionnaires; demographic and contextual data; and other student information.

The initial vectors concatenate this student contextual information into a series of identifying "contrasts" comprising the following variables and interactions.

• Categorical variables are derived by expanding the response options of a questionnaire variable into a binary series of one-degree-of-freedom "dummy" variables or contrasts. These form the majority of variables in the vector.
   
  
• Demographic variables possess ordinal response options, such as number of hours spent watching television, which are included as linear and/or quadratic multi-degree-of-freedom contrasts. Starting in the 2002 and 2003 assessments, and continuing since then, linear multi-degree-of-freedom contrasts were replaced by binary series of one-degree-of-freedom "dummy" contrasts, and quadratic multi-degree-of-freedom contrasts were removed.
   
  
• Continuous variables, such as student enrollment in a school or in a grade within a school, are included as contrasts in their original form or a transformation of their original form.
   
  
• Interactions of two or more categorical variables form a set of orthogonal one-degree-of-freedom dummy variables or contrasts.

For each variable in the population-structure model, there is a "missing" category. For example, for the gender variable, the categories are male, female, and missing. "Missing" is coded as one level of a variable in the model.

The specifications used for constructing the variables used in the scale score distribution models are provided, along with a summary. Starting in 2008 and continuing forward, the listing of estimation variables used in the population-structure models are presented in textual format. For more information about NAEP variables, see lists of each variable available by assessment year and subject.

Printer-friendly Version