Fundamental principles of epidemiological data analysis

This paper provides the statistical tools for analyzing simple epidemiological data. The term simple data refers to the most elementary data that can be obtained from an epidemiological study, such as crude data from a study with no confounding factors. Because the emphasis is on estimation, as opposed to statistical significance testing, the paper concentrates on formulas for obtaining confidence intervals for basic epidemiologic measures, although formulas which derive p values are also included. The formulas presented here give only approximate results and are valid only for data with sufficiently large numbers. When emphasis is placed on whether a confidence interval contains the null value (thereby converting the confidence interval into a statistical test), it may appear to matter if the limit changes its value slightly with a different formula and the limit is near the null value, a situation equivalent to being on the borderline of statistical significance. Placing emphasis on the exact location of a confidence interval, that is, placing emphasis on statistical significance, is an appropriate but potentially misleading way to interpret data. The precise location of a confidence limit can be ignored, and instead, the general width and location of an interval considered.