Self-selection bias occurs when individuals select themselves into a group, causing a biased sample with non-random characteristics. This can significantly affect the validity of epidemiological studies, as the participants who choose to be part of the study may differ in important ways from those who do not.