Ideally, an experiment should have an independent variable, a dependant variable, and a control variable. Researchers take steps to eliminate as many extraneous variable as possible, although eliminating all of them might be impossible in certain scenarios.
Experiments generally seek causation, and keeping the total number of variables as low as possible is essential for doing so. However, it is sometimes possible to rule out the effect of extraneous variables using statistical methods. For medical experiments, this is often necessary. These techniques, however, may call an experiment's results into question.
When performing exploratory research, however, scientists often examine as many variables as possible. When determining if a particular drug has detrimental side effects, it is important to look at as many possible effects as the data allows because ignoring potential variables can cause a dangerous side effect to go unnoticed.
Finding a correlation in an exploratory experiment does not prove causation, and one of the hazards of looking at many variables is interpreting coincidence as causation. When examining a large set of variables, correlations are likely to arise simply based on statistical noise. While these results do not show causation, they do give researchers areas for future experimental research to determine if the apparent effect is real or just a statistical fluke.