Test Hypotheses

How To... > Work with Statistical Objects > Test Hypotheses

You’re working with a collection that can be treated as a random sample from some population. For example, it could be a sample of students from a school, data from an experiment about animal learning, or a sample of census data from a state. You have a hypothesis you wish to test, such as test scores for students in the school have risen above some threshold, animal learning works better with rewards than it does with punishments, or median incomes for young workers measured in constant dollars has increased in the last ten years. These hypotheses are about the population, but your data are for a sample.

The statistical inference process can help you decide whether the measurements in the sample can be explained by chance variation as opposed to your hypothesis.

Note: To do a paired t-test, create a new attribute containing the difference between the two elements, and perform a hypothesis test to see if the mean difference is zero (see Define an Attribute by Formula Using the Collection’s Inspector).

Fathom ships with four sample documents that show how to perform non-parametric tests (see Non-parametric Tests).