Deciding on a test

Deciding on a statistics test eventually becomes easy. However, at the start it can seem quite daunting. There's a simple flowchart you can download to help with the process. In most cases, especially if you use the chart to plan your test BEFORE you collect your data, then you will be able to keep statistics nice and simple. You may not know about variances and normal distributions before you collect all your data, but at least thinking about your tests should mean you can narrow down your choice to two or three tests.

We recommend that the first few time you run a test, you run through from the 'Comparing means' or 'Examining relationships' sections of the website (see links on the left). After a few runs through with different data, you may want to jump straight to the specific test.

Types of decision you need to make

The independent variable. This really is key. If your independent variable is categorical, you compare averages of each category. If the independent variable is continuous, you look for relationships in your data.

The number of categories. There are special cases and different tests if you only have two categories or levels in your (categorical) independent variable.

Are distributions normal?. You've probably seen this from exploring your data already. The flowchart is quite vague, and asks if there are major deviations from normality. If there are, we use 'non-parametric' statistics. However, this doesn't completely resolve issues with non-normality of data, but is the best option in many cases. Minor deviations, especially from low sample sizes, can often be overlooked.

* The asterisk in the flow chart is there to indicate, in many cases it is easiest to assume equal variances and normality initially, as it is often easiest to test these assumptions after conducting the 'parametric' test (i.e. ANOVA or regression). If they are subsequently violated, you go back, and follow the route for 'no'

We recommend you follow through the tutorial using the buttons below. Once you have done this, you can subsequently use the navigation menu on the left to find the test you want.

Back Next