How to Do a t-Test in Fathom

Posted on May 3, 2012

1


How do you do a regular old t-test, no randomization? It’s easy, but different from how you do it in other programs.

1. Make a “test” object. Look at the shelf (a.k.a. toolbar), find the “test”, and drag one into your document.

2. Choose the kind of test you want from the menu. Here, we’re choosing “Test Mean.”

Image

The test changes to a test of the mean. It looks like this:

It contains default summary statistics: n=20, x-bar=0.5, s=1.0, yielding a p-value of 0.038.

How do you change it to what you want?

First of all, you have a choice:

  • You can do the test using summary statistics (N, x-bar, s), or
  • You can do the test using raw data

If you’re using raw data, simply drag the name of the variable (e.g., from a column heading in a table) to the little strip at the top where it says Attribute (numeric): unassigned.

Fathom reads the data in the column and does the appropriate computations:

Even now it doesn’t look like what you may be used to. But all the information is there, down to the p-value (0.039) at the bottom. To make it look more familiar, go to the Test menu at the top of the screen and un-check Verbose:

By default, Fathom does a two-sided test against a null value of 0.0. To change that, edit the blue text. Just go into the box and edit it like any word-processing text. You can change the value 0 to anything you like; and the phrase “is not equal to” is actually a menu.

If you don’t have the original data—no column of numbers, just the summary statistics—start with a new test and use the same edit-the-blue-text principle to enter the sample mean and standard deviation and the sample size, and alter the test’s null value and direction to fit your situation.

Advertisements
Posted in: How-to