On page 289 there is a box titled “Premises need for a good generalization”:
- The sample is representative
- The sample is big enough
- The sample is studied well
There should also be a small margin of error for the population sample (about +/-2%) and a larger confidence level for the population sample (about +/-95%). Both give strength to the generalization given and have more reason for it to be accepted.
Variation is also important in a population. If the number of people who own motorcycles is being tested for the state of California, the sample should be testing a more or less equal amount of men and women.