Table 1

Characteristics of cross-validation and external validation analyses for a new microsimulation model of smoking behaviours and outcomes

Analysis	STOP-generated output of interest	Comparator	Measure of goodness of fit
Cross-validation*	1950 birth cohort prevalence of never, current and former smokers, ages 0–70 years, by sex	CISNET-modelled 1950 birth cohort prevalence of never, current and former smokers, ages 0–70 years, by sex	RMSE
External validation: mortality	Cumulative mortality by age, sex and smoking status	Mortality rates of 1997–2009 NHIS respondents by age, sex and smoking status, and cumulative mortality	MAPE and RMSE
External validation: smoking prevalence	Prevalence of never, current and former smokers, annually, 1998–2009	NHIS never, current and former smoking prevalence, annually, 1998–2009	RMSE

*The initial cross-validation did not include smoking relapse. In the subsequent calibration step, we incorporated smoking relapse and calibrated cessation to achieve a good fit to the CISNET-modelled 1950 birth cohort prevalence of never, current and former smokers.
CISNET, Cancer Intervention and Surveillance Monitoring Network; MAPE, mean absolute percentage error; NHIS, National Health Interview Survey; RMSE, root-mean-square error; STOP, Simulation of Tobacco and Nicotine Outcomes and Policy model.