Talking about teachers: Would sampling weight adjustments allow for teacher-centric inferences in future TIMSS assessments?

Abstract

The Trends in International Mathematics and Science Study (TIMSS) is an influential mathematics and science assessment administered at fourth and eighth grades in around 60 countries across the world. Under current practices, TIMSS does not support claims about teachers unless the inference is drawn as a student characteristic (“Half of the eighth graders in the country have mathematics teachers with Master’s degrees.”). With this in mind, this paper proposes and tests through a simulation study two methods that would allow for teacher-centric inferences (“More than half of the eighth-grade mathematics teachers have Master’s degrees.”) if links between sampled teachers and the unsampled yet eligible classes were known. The first method posits that with such links inverse-probability weights could be calculated, and the second method, the Multiplicity-Adjusted Indirect Sampling (MAIS) approach, posits that teacher estimates could be obtained by adjusting downward the classroom weights based on these links.