Award Date

5-1-2012

Degree Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Educational Research Cognition and Development

First Committee Member

Gregory J. Schraw

Second Committee Member

W. Paul Jones

Third Committee Member

Kendall Hartley

Fourth Committee Member

Peter B. Gray

Number of Pages

161

Abstract

Typically in calibration research, subjects perform a task and make a judgment about the success of the task. Accurate findings help subjects improve self-calibration. In addition, researchers rely on the accuracy of findings to make inferences about underlying metacognitive processes. Consequently, it is important that the measures used to assess monitoring accuracy are as free of bias as possible. Bias indicates whether an observed value of monitoring accuracy over- or underestimates the true value.

This study examined gamma and three other viable statistics, d', C, and the G Index, currently used to measure monitoring accuracy. Using Monte Carlo simulation techniques, I sought to determine whether bias occurred in the use of any of the four measures, and if so, which was least biased. I found substantial bias in gamma, d', and C, especially in shorter and less difficult experimental conditions. Bias was caused primarily by the presence of empty cells in a 2 x 2 contingency table. Empty cells, in turn, were caused by a small sample size that was often insufficient to populate all the cells in the table with at least one response. The fourth statistic, the G Index, was not affected by the same computational issues as the other statistics and showed no bias.

Keywords

Bias; Calibration; Discrimination; Gamma; Metacognition; Metamemory; Objectivity; Self-control; Self-regulation

Disciplines

Cognition and Perception | Educational Psychology

Language

English


Share

COinS