ERIC Number: ED064393
Record Type: Non-Journal
Publication Date: 1972
Reference Count: N/A
Jackknifing Disattenuated Correlations: National Assessment of Educational Progress.
Rogers, W. Todd
The jackknife is a general inferential technique intended to ameliorate the problems associated with inadequate sampling theory. The research reported herein is directed at investigating the utility of the jackknife for establishing confidence intervals on and testing hypotheses about the disattenuated correlation coefficient for small samples. A review of the literature is made. Several computer simulations were performed to investigate the utility of the jackknife. The theory of the jackknife implies that the jackknife statistic is approximately distributed as a Student - t variate with the appropriate degrees of freedom. Results include: (1) The direction of the difference between the theoretical and actual cumulative proportions of jackknife statistics which were at or below the 10 percentile points of comparison varied across the different values of p(TX, TY); (2) The jackknife was sensitive to changes in the values of the reliabilities for each combination of p(TX, TY) and N;. Conclusions include: (1) It is unlikely that a mathematical model can be formulated to describe the sampling distribution; (2) The performance of the jackknife was sensitive to changes in the values of the input a parameters; and (3) The jackknife can be used to set approximate confidence intervals. (CK) Aspect of National Assessment (NAEP) dealt with in this document: Procedures (Sampling).
Descriptors: Computer Science, Correlation, Educational Programs, Hypothesis Testing, Models, Problem Solving, Program Evaluation, Sampling, Simulation, Theories, Values
Laboratory of Educational Research University of Colorado, Boulder, Colorado
Publication Type: N/A
Education Level: N/A
Authoring Institution: Colorado Univ., Boulder. Lab. of Educational Research.