This paper studies the effect of different epidemiologic designs on measures of association between disease outcome and exposure variable in a table and obtains maximum likelihood estimators, as well as their sample variances, for six measures of association under threemain types of epidemiologic study designs (Hsieh et al. [7]). We investigated the properties of the estimators so derived with the following main findings: (1) For relative risk (RR), odds ratio (OR) and attributable risk fraction (ARF), the asymptotic variances and the asymptotic confidence intervals are the same for all designs while for population attributable risk (PAR) and population attributable risk fraction (PARF), the asymptotic variance for cross-sectional study is greater than those for cohort and case-control studies; (2) theasymptotic variances for RR and particularly OR have slow convergence rate while the asymptotic variances for ARF and PARF have good convergence rates; (3) the percentile intervals for cross-sectional study are wider than those for cohort and case-control studies; (4) the differences in variances and percentile intervals decrease among the three designs as the sample sizes increase.
