A video game developer wants to know how long it takes people to finish playing a new game.
She surveyed a two random samples of 20 players and asked how each long it took them (in minutes).
How could the developer change her samples to make you more confident that the data is representative of the population?