Associations between risk factors and obesity in childhood using multivariate conditional logistic regression model, 1996–2006.

Data aged 2–7 years from the 9 cities was used for analysis, with 1016 pairs in 1996 and 1593 pairs in 2006, because of the full coverage of factors surveyed in this age range.Conditional logistic regression was used to screen potential risk factors by stepwise selection and 0.05 for sizes of a test.aβ represented parameter estimate with positive value as possible risk factors and negative value as possible protective factors.bOR, Odds ratio; CI, Confidence interval.cBreastfeeding includes exclusive and partial with 〉1/2 of total intake.dBody mass index (BMI) is calculated as weight (kg) divided by height (m) squared through self-reported weight and height.eRunning includes running, jumping, climbing stair, et al; sitting represented not or hardly playing when others doing.