Some Methodological Issues in Health Psychology Research

Abstract

Four issues of a methodological nature that are perceived to be recurring problems in research directed at predictors of illness are pointed out. In the first place, sample size affects statistical power (i.e. the probability of obtaining significant results). Methodological literature tends to caution against the use of samples that are too small, but tends to ignore the problems arising from obtaining statistical significance for trivial effects due to samples that are very large. Possible solutions to this problem are mentioned. Secondly, the reduction of criterion variability due to the operationalization of the criterion variable may result in failure to find statistical significance. This problem is of special relevance when patients already suffering from a particular illness are used and survival rates are used in lieu of the degree of that illness. Thirdly, the demand for a more differentiated set of hypotheses is emphasized. This situation arises because of the non-experimental nature of research and the consequent multivariate nature of research problems in this area. Elementary statistical procedures do not allow for theoretically derived hypotheses that are really meaningful, to be investigated adequately. However, using more advanced procedures, such as path analysis and structural equation modelling, not only requires a sound theoretical background (in health psychology) but also a knowledge of these procedures. Finally, the interrelatedness of statistical design, measurement and statistical analyses is emphasized. Because of this interrelatedness, the delegation of research design to statisticians, who cannot be expected to be familiar with theoretical issues and measurement possibilities in the field, is bound to result in impoverished research findings. As a result, a sound grounding in methodology is of great importance to a burgeoning field such as health psychology.