Influence of research design upon data analysis

Considerations regarding the interrelationships among theory, research design, observational approaches, and data analysis techniques bear on the consistency and integration of our overall procedures in performing social research. It is here illustrated how an initial research conclusion, shown through further analysis to be false, might have been avoided entirely if a more stringent research design had been employed. In a medical treatment context a worrisome negative correlation between treatment and outcome is revealed as spurious, due to gender differences in reactions to a serious disease. If subjects had been randomly assigned to experimental and control groups, a zero correlation coefficient would have correctly indicated that the treatment had no effect upon the outcome.