Table 5 The generalized estimating equation (GEE) of BMI for six-year surveys retaining significant exploratory factors

Variable Category Estimate 95% confidence interval P value P value in type 3 effect analysis
BMI at baseline survey   0.151 0.141 0.161 <.01* <.01*
Age   -0.263 -0.268 -0.258 <.01* <.01*
Overweight at age 5 ½ years   1.742 1.710 1.775 <.01* <.01*
(Base category; < 85th percentile)
Gender (base category; boy) Girl -0.160 -0.181 -0.139 <.01* <.01*
Gestational period < 22–36 weeks 0.160 0.106 0.214 <.01* <.01*
(Base category; < 37–41 weeks) ≥ 42 weeks -0.144 -0.262 -0.003 <.01*
Main parental care provider at age 6 months Father -0.107 -0.389 0.175 .46 <.01*
(Base category; mother) Grandmother 0.024 -0.038 0.087 .44
Grandfather 0.013 -0.321 0.347 .94
Others 0.122 0.063 0.181 <.01*
Main parental care provider across the surveys Father -0.039 -0.186 0.108 .60 <.01*
(Base category; mother) Grandmother -0.055 -0.100 -0.010 .02*
Grandfather 0.108 -0.053 0.268 .19
Others 0.087 0.071 0.102 <.01*
Wake-up time (base time; 0 am)   -0.044 -0.062 -0.026 <.01* <.01*
Video game playing < 1 hour 0.014 -0.010 0.037 .25 .02*
1 hour - 0.068 0.029 0.107 <.01*
2 hour - 0.009 -0.104 0.122 .88
3 hour - -0.020 -0.265 0.225 .87
Snack at age 4 ½ years Eating irregularly -0.047 -0.083 -0.010 .01* .03*
(Base category; eating regularly) Don’t eat -0.044 -0.148 0.059 .40
1. *Statistically significant at the level of P < .05.
2. BMI: body mass index.
3. Type 3 effect test is based on an asymptotic chi-square distribution of the likelihood ratio statistic computed from GEE models and it allows one to test for an effect of interest provided all the other variables are included in the GEE model.