Unlike experimental studies where the researcher has a strong degree of control over sources of bias and variation, observational studies present major hurdles to conducting research that provides strong clinical evidence. Issues such as missing data, confounding bias and effect modification mean that we need to use statistical modelling to adjust or offset these problems.

In this series, biostatistical models used for the analysis of binary clinical outcomes will be discussed.

• Measuring effect size for binary outcomes: Risk difference, relative risk and odds ratios
• Binary logistic regression
• Communicating your results