Related Experiment Videos
Risk behavior data analysis: ordinal or dichotomous the choice is yours
J Wanzer Drane1, Robert F Valois
1Department of Epidemiology and Biostatistics, Arnold School of Public Health, University of South Carolina, Columbia 29208-0001, USA. WDrane@gwm.sc.edu
Objective:
To demonstrate the differences of 2 approaches to data analysis.
Methods:
Using the South Carolina YRBS data, study focused on contingency tables and ANOVA. Additive chi squares are utilized to illustrate information loss when collapsing a contingency table. Odds ratios are derived from contingency tables or logistic regression. Means are utilized in ANOVA. Five measures of life satisfaction were summed to create a pseudo-continuous response variable that was subsequently trichotomized. All predictors are dichotomized risk variables.
Results:
Chi squares from subtables added exactly to that of the original table measuring lost information. ANOVA conveyed the same clinical message.
Conclusion:
Clinically relevant conclusions might be the same even when drawn from any of several different analyses of the same risk-behavior data.