ERIC Number: ED204353
Record Type: RIE
Publication Date: 1979-Sep
Reference Count: 0
Convergence and Robustness of Analytic Techniques for Qualitative Data. Final Report, July 1, 1978 to September 30, 1979.
Borgen, Fred H.; Muchinsky, Paul M.
The purposes of this study are to compare alternative statistical techniques for analysis of the kinds of qualitative data commonly obtained in educational settings. Distinctions between chi-square based measures for contingency tables and proportional reduction of error measures of the strength of association are reviewed, and it is argued that there are direct equivalencies between the chi-square based measures and at least one of the proportional reduction measures, Goodman and Kruskal's tau. A direct arithmetic linkage between tau, chi-square, and Cramer's V is demonstrated, and the implications of this linkage are discussed. In order to study the robustness of these various analytic techniques, the Monte Carlo simulation technique was used to construct 45 different contingency tables, varying in sample size and levels of Cramer's V. Power differences for chi-square and tau indicate that underlying relationships are many times more likely to be detected by chi-square than by tau. The final question of this study was the robustness of chi-square tests when the contingency table marginal totals were significantly skewed. The power decrement was about 10-15% when the skew was extreme; the adverse effect of moderate skew was quite mild. (BW)
Publication Type: Reports - Research
Education Level: N/A
Sponsor: National Inst. of Education (DHEW), Washington, DC.
Authoring Institution: Iowa State Univ. of Science and Technology, Ames.
Identifiers: Chi Square; Coefficient of Association; Contingency Tables; Robustness