ERIC Number: ED095217
Record Type: Non-Journal
Publication Date: 1974-Apr
Reference Count: N/A
A Comparison of Interval Estimation of Coefficient Alpha Using the Feldt and the Jackknife Procedures.
Pandey, Tej N.; Hubert, Lawrence J.
This investigation had two major purposes. The first was to explore the use of an inferential technique called Tukey's Jackknife in establishing a confidence interval about cooefficient alpha reliability. The second purpose was to study the robustness of the Feldt and the jackknife procedures when the data fails to satisfy usual normality assumptions. Using a linear model of test score, computer simulated data representing a matched item-examinee sample from their respective populations was employed. The two jackknife procedures, is this situation, were less conservative. For log-normal distributions of examinee score components, Feldt's procedure gave reasonable interval estimates for low values (=.60) of population alpha, but for higher values of population alpha (=.90), the Feldt procedure was not found to be robust. The two Jackknife procedures, on the other hand, were found to be relatively robust over the entire range of investigation of values of alpha. None of the procedures was found useful in interval estimation of coefficient alpha when the distribution of examinee score components were of extreme double-exponential form. (RC)
Publication Type: Speeches/Meeting Papers
Education Level: N/A
Authoring Institution: N/A