TG gave a great explanation. One thing that’s worth adding, though: Testing hypotheses requires you to isolate the factor(s) being tested. In TG’s affirmative action example, you would need to identify the effects of other causative factors and (if possible) either eliminate those extraneous influences or somehow “factor out” their effects. For example, minority enrollment in college could also drop from one year to the next if the real number of minority high-school seniors dropped, or if the percentage of minority high-school seniors relative to all high-school seniors fell.

It’s a lot like filtering out noise to get a clean signal…

This is the process of generating an idea from observation and then testing it to determine its validity. For example, one might hypothesize that companies which cut costs by downsizing their workforce would generate more profits and thus a higher stock price. You could test this by looking at companies which have done just that, and comparing changes in their stock price over time against similar companies who have not made significant cuts.

You can do the same thing in other environments. In the workplace, for example, some companies have combined vacation and sick leave into general leave time, in a belief that this would reduce absenteeism. Easy to test. In education, a number of states have outlawed affirmative action, saying that this will actually increase the numbers of minorities going to college. Another easy one. In politics, consultants do this all the time. They tell a candidate to, for example, come out against ilegal immigration and their poll numbers will go up. Easy to track.

Hope this helps.

this might help if its a statistics class, otherwise the previous answers are perfect.

http://en.wikipedia.org/wiki/Statistical_hypothesis_testing