Loading Events

Dan Black (Univ Chicago)

17 September 2013 @ 12:00 - 13:15

 

  • Past event

Details

Date:
17 September 2013
Time:
12:00 - 13:15
Event Category:

“Duke Grads, Monkeys, and Jobs and Wall Street: The Use and Misuse of Latent Variables”

abstract

We consider common Item Response Theory (IRT) measures of latent variables and consider their use as independent variables in regression analysis. We show that because of the inherent measurement error in the construction of IRT scores that OLS estimates are necessarily biased. Because of the structure of the measurement error, standard instrumental variables estimation fails to produce consistent estimates. Moreover, when latent proficiency is an independent variable, we show that the standard institutional “Plausible Values” methodology produces biased inference because the institutional conditioning model places restrictions on the form of the secondary analysts’ model. We offer an alternative approach that avoids these biases based on the mixed effects structural equations (MESE) model of Schofield (2008). While we develop our arguments with standardized test scores, our analysis extends to any latent measure.