Aspects of the design protocol and the statistical methods for analysis of tar, nicotine and carbon monoxide yields in cigarette smoke that can affect the measurement variability within collaborative studies.