ANALYSIS WITH INCOMPLETE DATA: A MONTE CARLO EVALUATION OF INTERVAL ESTIMATES UNDER SPECIFIED CONDITIONS OF SELECTION.
Item
-
Title
-
ANALYSIS WITH INCOMPLETE DATA: A MONTE CARLO EVALUATION OF INTERVAL ESTIMATES UNDER SPECIFIED CONDITIONS OF SELECTION.
-
Identifier
-
AAI8319789
-
identifier
-
8319789
-
Creator
-
PERRY, PHILIPPA CARRINGTON.
-
Contributor
-
Alan L. Gross | David Rindskopf
-
Date
-
1983
-
Language
-
English
-
Publisher
-
City University of New York.
-
Subject
-
Education, Educational Psychology
-
Abstract
-
The present research examines estimates of the relationship between two variables derived from samples where selection has taken place. Paired information is present only for certain selected subjects and, from such incomplete samples, inference is to be made for the relationship in the total group of applicants. We explore the extent to which selection conditions and violations in the underlying assumptions for the data structure affect the value of the estimates derived.;In particular, a vector of test scores (Xt) is available for Nt subjects. Through a selection process, criterion scores (Y) are available for Ns selected subjects, Ns < Nt. For Nt - Ns cases, however, Y scores are missing. We want to estimate the XY relationship in the full applicant group, the total sample. Estimates for the XY relationship in the total sample are derived from a probability distribution of the missing Y scores. These estimates for regression coefficients, residual variance, and difference coefficient, have been derived under restrictive conditions for the sampling distributions and selection modes. This research evaluates the accuracy and sensitivity of these estimates when the underlying assumptions are violated.;A 4 x 3 factorial design is created setting 4 distributional types against 3 selection modes. At least 100 simulated samples are computer generated for each distribution-selection condition and expected values are computed for each estimate.;Cell (1,1) investigates the accuracy of the proposed interval estimates when the assumptions of the model are met; the remaining 11 cells explore and compare the effects of separate, as well as simultaneous, violations of distribution assumptions under each specified selection mode. An estimate for the total sample correlation is computed and evaluated as well.
-
Type
-
dissertation
-
Source
-
PQT Legacy CUNY.xlsx
-
degree
-
Ph.D.
-
Program
-
Educational Psychology
