Chapter 2 Statistical Hypothesis Testing _________________________________________________________________ _
that rejects the null hypothesis at 0.05 level of significance is called
while a test that rejects the null hypothesis at 0.01 level of significance is called
Step 3. Select the Appropriate Test Statistic
A test statistic is a value, determined from sample information, used todetermine whether to reject the null hypothesis. We will deal with these in detailslater. Now consider the tossing of a coin problem posted earlier. Since we wantto test the “fairness” of the coin, the appropriate test statistic is the number of heads occurred in
= 100. We can also use the ratio between thenumber of heads occurred and the total number of tosses.
Step 4. Formulate the Decision Rule.
A decision rule is a statement of the specific conditions under which thenull hypothesis is rejected and the conditions under which it is not rejected. Theregion or area of rejection defined the location of all those values that are so largeor so small that the probability of their occurrence under true null hypothesis israther remote.Going back to our problem regarding the “fairness” of a coin, note that if the null hypothesis is true (i.e.,
= 0.5), then the number of heads in 100 tossesshould not be too different from 50. So if the decision rule is to reject
if andonly if 47
, then we tend to reject the null hypothesis very often.This is the same as saying that a defendant is guilty if one witness appears in thecourt! On the other extreme, if we reject the null hypothesis if and only if
, then we can think of this as having a roomful of witnesses before wesay that the defendant is guilty. In practice, the decision rule will be based on thevalue of
and the probability distribution of the test statistic.
Step 5. Make a Decision
The fifth and final step is computing the value of the test statistic, check whether its value is inside the rejection region, and making a decision to reject or not reject the null hypothesis.
P-values in Hypothesis Test
A commonly used method in testing hypothesis is to report the
of thedata. A
value does not require imposing a pre-selected level of significance.(Sometimes, selecting the appropriate
is troublesome to the researcher.) The
-value is the probability that the test statistic will take on a value that is at least asextreme as the observed value of the statistic when the null hypothesis is assumedto be true. Thus, a
-value conveys much information about the weight of evidence against
and so a decision maker can draw a conclusion at anyspecified level of significance. Formally, we define the
the smallestlevel of significance that would lead to the rejection of the null hypothesisgiven the observed data.
With this definition, if the observed
-value is equal to0.40, then we will reject the null hypothesis at this level. But using this value istoo large! We can’t afford to have this very large probability of committing type Ierror. Thus, we tend to do not reject the null hypothesis. On the other hand, if
-value = 0.0001, then we can reject the null hypothesis at this very small level of 19