chap10 Fritz
Section 10.3 Notes

I. Using significance tests

A. Choosing a level of significance
1. asks yourself, "how much evidence is required to reject the null hypothesis?"
a) the more plausible the null hypothesis, the more evidence you need, thus a smaller p-value is required
b) also depends on the consequences of such a decision
2. There are no sharp boundaries between significant & not significant, ONLY increasingly stronger evidence as the p-value gets smaller.

B. What statistical significance doesn't mean
1. Even though the null hypothesis can be rejected & an effect is present, the effect may be extrememly small
2. Rejecting the null hypothesis does not mean that the effect is strong
3. Look at the distribution of the data & look for any outliers

C. Don't ignore lack of significance - this is just as important as if you reject the null hypothesis

II. Abuse of tests

A. Statistical inference is not valid for all sets of data
1. hypothesis tests cannot correct flaws in experimental design
2. Must have randomization - tests & intervals are based on probability laws
B. Beware of searching for significance
1. remember that an extremely large sample size can cause something to be significant
2. also be wary of p-values that are close to the alpha level - not much evidence to reject or fail to reject the null hypothesis
Last updated  2008/09/28 08:17:31 PDTHits  320