Why do hypothesis need to be tested?

Answer 1

It is a simple question, but sometimes simple questions do not have simple answers.

I have included two related links which I feel are very helpful. Don't be worried if you don't understand much of the math in the second link. You may find less mathematical explanations by searching the internet for "hypothesis testing"

You asked this both in science and statistics. You know sometimes the same word can have two meanings. Hypothesis testing is one of them. I will explain why we must test hypotheses in science, and what it means to test them in statistics.

In science, a hypothesis is a speculative idea or explanation of a phenomena. Evidence or data is collected in an unbiased manner as possible to either prove it or disprove it. But how much data or evidence do we need? Sometimes, our hypothesis becomes a theory, a reasonable explanation that seems to fit circumstances or events, that will help us make decisions. As more observations seem to support the theory, we consider it to be valid or truthful theory. Many, for example, consider global warming to be a valid theory.

Now, for the usage in statistics. Hypothesis testing is a statistical method. Hypothesis testing tells me if I have sufficient data to draw a conclusion, given a certain level of significance.

I will give you an example:

I have gathered some data and calculated a statistics on smoking. I found in my sample more women smoked than men. But, of course I didn't survey everyone, so there is a chance that my data has error in it, and perhaps I really don't have the necessary support to make this statement about everyone (the general population). So, I use a statistical test, with one hypothesis contrary to what my data suggests, that women and men smoke equally, which we call the null hypothesis. Now, I have a second hypotheis which we call the alternative hypothesis, which is that women smoke more than men. To complete the test, I need to include an "alpha factor" or the level of significance. I can with this factor, make it very easy to disprove the null hypothesis or very difficult. I generally use this factor to make the criteria for choosing between two hypothesis consistent.