Because the children were sampled from 2 different populations,
all analyses were calculated independently for each sample.
Prior to our primary analyses, we examined weekday–
weekend day effects by using multilevel analysis of variance,
which showed significant heterogeneity (defined a priori as P
0.10) between weekend days and weekdays in energy intake and
percentage of energy from fat with the television on or off in both
samples (F ranged from 3.57 to 15.16, all P 0.06). Therefore,
all subsequent analyses were performed separately for weekend
day and weekday data, or effects for weekend day compared with
weekday were included in the analyses. The average of the 2
weekdays were used in all weekday analyses.