Blackwell Reference Online

Extract

The history of hierarchical linear modeling (also called multilevel modeling) can be linked to the seminal work of Robinson (1950) in recognizing contextual effects. The theory that Robinson illustrated is also sometimes referred to as the “frog/pond” theory and bears discussion here. Suppose that a biologist has two frogs lying on an examination table that both weigh 500 grams and are both 1 year old. Consider further that Frog A was drawn from Pond A where it was the largest 1-year-old frog in the pond, and Frog B was drawn from Pond B where it was the smallest 1-year-old frog in the pond. The possible disregard of the contextual effect of the frogs' habitat could lead to some erroneous conclusions about frog development if the two frogs were considered outside of the pond from which they were drawn. Placing this same contextual argument in educational terms, let us suppose that a researcher is interested in monitoring reading proficiency among students within New Mexico's schools. In testing differences among the schools, the researcher might use an analysis of variance (or ANOVA) to test for differences between mean reading proficiency scores across schools. As was noted with the frog/pond illustration, doing so would neglect the fact that some of the schools closer to the border of Mexico might have a larger number of non-English speaking students. Neglecting this structure ... log in or subscribe to read full text