Library News and Blogs

How to Choose the Correct Statistical Test Down to a T (-test): a Real-Life Example

This blog was written by Hannah, a third year PhD student in Psychology and a Stats@Liverpool tutor. Hannah can help with a range of statistics issues such as understanding basic statistical concepts, choosing the most appropriate statistical test, analysing, interpreting and visualising data (predominately using SPSS with a basic understanding of R) and more. 

One of the most common statistics problems that undergraduate students face is how to choose the correct statistical test. This is particularly relevant when students reach their dissertation year and are often left to their own devices. For many students, planning their analysis is pushed to one side as they do not know where to start.   

This blog aims to help students decide on how to choose the correct statistical test for their work. It also includes an example of a real session that I had with a student (Student X) and the tips and tricks we use to scare those statistical worries away!

What does a Stats@Liverpool Session look like?

Student X booked a Stats@Liverpool appointment with me after collecting their data for their dissertation. They had a SPSS file filled with hundreds of numbers and staring at it made them feel rather overwhelmed. The first question they asked is, “what do I do with all this data?”. And that is when the session started.

During the first part of the session, we took a step-back to talk about the data itself. I explained to the student that in order to choose the best statistical test for their data, they had to ask themselves a few questions first:   

1. What are your research questions/hypothesis? Simply, what are you trying to find out?
2. Are you looking for a test of difference (e.g., is one group different from the other?) or a test of relationship (e.g., does one variable predict an outcome?)?
3. What is your independent variable(s) (the one(s) you change/manipulate) and dependent variable(s) (the one(s) you measure)?
4. What type of data are your independent and dependent variables? Are they scale/continuous (e.g., can they be any number?), ordinal (e.g., do they have a category with an order, like a shoe size?), or are they nominal/categorical (e.g., can you put them into categories with no order, like gender?).
5. Is your data between-subjects (e.g., do different participants take part in different conditions?) or within-subjects (e.g., do the same participants take part in more than one condition?)?
6. If it's between-subjects – how many different groups do you have?

Note: don’t worry if you do not know the answers to all these questions – that is what we are here for!

Once I discussed the above questions with Student X, it was time to think about the statistical test. You may be wondering at this point how answering all these questions are going to help? Luckily, there is something called a statistical analysis tree, which we often use at Stats@Liverpool to help students (see picture).

The tree follows a similar pattern as the above questions and helps guide you to the right statistical tests for your data. Using this, I was able to help Student X decide what statistical test they wanted to do (remember: we cannot tell you what test to do - especially if it’s an assignment!).

The outcome of the session

After the session, Student X emailed me to say they were feeling much more confident with their analysis – feedback that is always nice to hear. They said that they hadn’t heard of the statistical analysis tree before and will probably use it in the future (disclaimer: I still use it sometimes too - even as a PhD student!). Sometimes, talking through your data with someone (even if you think you know what you are doing) can give you all the reassurance you need.

If you want any help with choosing the correct statistical test (or other stats issues) then please book a session with me, or another Stats@Liverpool tutor or find out more about our online sessions.