Testing Hypotheses

Also called **Test of Significance**

Only reject or not reject, there's no accept hypotheses.

Step 1: construct *null hypothese*s, and *alternative hypotheses*

*null hypotheses*- null means have no value. We assume it right, and going to reject it if the data support this.
*alternative hypotheses*- results which we want

Step 2: in t-test example, to test difference between two means, calculate the difference, standardize it to 1 by dividing it by standard error. So the result t should be t-disruibuted. If the t is located very near the tails, probability very low, that is, not significance.

Step 3: based on step 2, you can calculate *confidence interval*, do a specific significance test, or compare with p-value. (the three are the same in essence)

See also:

t-test

page revision: 5, last edited: 05 Sep 2009 08:33