Large studies have more statistical power than small studies. You can see the power of large numbers if you think about flipping a coin. Suppose you are trying to test the theory that a coin is fixed so that it lands “heads” more than “tails”. So you want to test whether it lands on “heads” more than half of the time, as would happen if the coin were fair. If you flip the coin twice and get two heads, you don't have a lot of evidence—it would not be unusual to flip a fair coin and get two heads in a row. With two coin flips, you can't say with much certainty if you have a fair coin or not. Even three or four heads in a row wouldn't be very unusual for a fair coin. If, however, you flipped the coin 20 times and got mostly heads, you would start to think that the coin is not a fair one.
With an increasing number of observations, you have more evidence on which to base your conclusions, so you have more confidence in them. It is a similar idea in epidemiologic studies. Say you are interested in finding out whether or not heavy alcohol use increases the risk of breast cancer. If there are only a few cases of breast cancer among the alcohol drinkers and the non-drinkers, you will not have much confidence in making any firm conclusions about the role of alcohol. If, however, there are hundreds of breast cancer cases, it is easier to draw firm conclusions because there is more evidence on—and, therefore, more confidence in—the link between heavy alcohol drinking and breast cancer.
Updated 03/13/09
