Sources of bias in a dietary survey of children