Value-added analysis uses statistics to estimate a teacher’s impact on student learning. But experts are still debating which value-added formula yields the most accurate ratings. To show how different models can produce different results, The Times calculated value-added ratings using four value-added models, represented here by different colored lines on the graph. Each model attempts to control for various factors beyond a teacher’s influence. In general, the results of all four models are very similar, though considerable variation occurs in some individual cases. On the advice of leading experts, The Times has elected to use Model 4, which adjusts for as many variables as was possible, given the information available.
Teacher rating page »
This model controls for each student’s past performance on standardized tests.
This model controls for each student’s past performance on standardized tests and student characteristics such as race, gender, English proficiency and parents’ education.
This model controls for each student’s past performance on standardized tests, student characteristics and “peer effects,” or the collective influence of all the students in the class.
This model controls for each student’s past performance on standardized tests; student characteristics; “peer effects,” or the collective influence of all the students in the class; and the prior performance of the entire class.
Math effectiveness
English effectiveness
Note: None of the models above controls for special education status, which the district withheld due to privacy concerns. For the same reason, the analysis used grade-level data on student race and ethnicity. A more detailed discussion can be found on p. 14 of the technical report.
Enjoy more stories like this for 