ERIC Number: ED334239
Record Type: RIE
Publication Date: 1991-Mar-21
Reference Count: N/A
Double Cross-Validation in Multiple Regression: A Method of Estimating the Stability of Results.
Rowell, R. Kevin
In multiple regression analysis, where resulting predictive equation effectiveness is subject to shrinkage, it is especially important to evaluate result replicability. Double cross-validation is an empirical method by which an estimate of invariance or stability can be obtained from research data. A procedure for double cross-validation is discussed, using heuristic and actual research data to illustrate a non-generalizable outcome and a generalizable outcome. The procedure involves the use of two samples or subsamples to produce two pairs of regression equations from which respective shrinkages can be determined. The more closely the shrinkage estimates approach zero, the greater the degree of stability across the subsamples and the more confidence the researcher can vest in the replicability of the results. The first example uses a heuristic data set of 20 (2 subsets of 11 and 9 subjects, respectively) with 5 predictors to illustrate that weights must be compared empirically rather than subjectively. In the second example, data are drawn from a study of the life satisfaction of 200 nursing home residents. Seven tables present study data. An appendix contains the Statistical Analysis System program used to analyze the data. (SLD)
Publication Type: Reports - Evaluative; Speeches/Meeting Papers
Education Level: N/A
Authoring Institution: N/A
Identifiers: Double Cross Validation; Invariance; Research Replication
Note: Paper presented at the Annual Meeting of the American Educational Research Association (Chicago, IL, April 3-7, 1991).