Testimate Guide

Configuration

This page describes the blue “configuration stripe.” Controls in the stripe let you set parameters for your test.

What appears in the stripe depends on the test. One simple example is configuring a test of mean. Here, we want to test whether the mean of pulse is above 70, but the test defaults to asking if pulse is different from zero:

The key is in the blue stripe at the bottom of the illustration. Click the ≠ button to change it to > and edit the zero to be 70:

The result is still “significant,” but applies to 70 rather than to zero!

We could also have changed the confidence level (currently 95%), which would have changed the output for the CI.

One- and two-sided tests

Clicking the ≠ button changes a two-sided test to a one-sided test. The button changes to > or < depending on whether the relevant statistic is greater or less than the value you’re testing against.

Special note: Configuring goodness of fit

A goodness-of-fit test operates on a single categorical attribute. You need to tell testimate the hypothesized proportion for each value of the attribute. The default is that all proportions are equal.

Using our pulse data with 16, 17, and 18-year-olds, let’s test to see if the three ages are equally represented.

In the configuration stripe, you’ll see a “disclosure triangle” labeled set hypothesized proportions. Click it to open up the control. You’ll see this:

With a \(P\) of 0.16, we can’t reject equal proportions; the numbers are different, but not different enough.

On the other hand, if we edit the proportion of 17 to be 0.4, we still can’t reject:

• When we edit the boxes under 17 or 16 to change their proportions, the proportion under 18 will react to make their sum equal to one.
• If you want to make all proportions equal again, press the equalize button.