We calculated all estimates using the appropriate sampling weight to represent US children aged 4 to 16 years. The purpose of the sampling weight is to provide population estimates that adjust for unequal probabilities of selection and account for nonresponse. The weights were poststratified to the US population as estimated by the Bureau of the Census. For analyses, we used two software packages (SAS, version 6; SAS Institute; Cary, NC; and SUDAAN, version 7; Research Triangle Institute; Research Triangle Park, NC [a program that adjusts for complex sample design when variance estimates are calculated]).16–17 Using logistic regression, we modeled factors predicting asthma severity, physician visits for asthma, hospitalizations for asthma, and FEV1 < 80% predicted, adjusting for age, for race/ethnicity, SES, family size, and parental history of asthma. Each model was evaluated for evidence of effect modification and confounding. For the evaluation of continuous lung function data (ie, FEV1, FEV1/FVC ratio, FVC, and MMEF), we developed linear regression models that adjusted for age, sitting height, sex, race/ethnicity, SES, parental history of allergy or asthma, family size, and cotinine levels. In addition, we used χ2 tests of trends in proportions (Epi-Info, version 6.04; Centers for Disease Control and Prevention; Atlanta, GA) to determine whether trends for asthma hospitalizations, physician visits, health status, FEV1 < 80% predicted, use of inhaled bronchodilators, and use of inhaled corticosteroids were significant across strata of increasing asthma severity and tobacco smoke exposure.