[On 'Negative control methods': related principles, methods and applications]

Abstract

Negative control methods (NCM) are developed based on the idea that negative controls should be tested with negative results. It also extends to the multiple comparison groups and used for specificity of association in population studies. Negative controls do not take part in studying the causal hypothesis but sharing the same potential bias structures. Under different purposes, NCM can be divided into methods on negative exposure, negative period or negative outcome controls. They are mainly used to detect and control bias, such as: selection bias, confounding bias and measurement bias, but leaving the sample size at secondary importance. Rational application of NCM can help improving the accuracy of causal inference in both experimental and observational processes.

Keywords: Bias; Causal association; Exposure; Negative control; Outcome; Period.