Using a random number generator to select participants.

Participants under-reporting their food intake due to social desirability bias.

Having a sample size of $n = 1000$ instead of $n = 500$.

Using a stratified design to ensure age balance.