statsART info

Latest News

Links


Glossary

When people first start learning about research methodology and statistical analysis, they often say that it is a bit like trying to learn a new language! It is true that there is lots of terminology, but it’s mainly long words used to explain something simple. This glossary is intended to help understand this strange language of statistics.

For your ease of navigation we have split the terms into three main categories selectable below:

Analysis of DifferencesAnalysis of RelationshipsAnalysis of Frequency Data

Independent variable: This is the variable that you manipulate in experimental designs. For example, if I wanted to know whether people feel happier after eating different types of chocolate, my manipulation would be the types of chocolate given to the participants. I might give people white chocolate, milk chocolate and plain chocolate to eat. In this example my independent variable would be the type of chocolate eaten.

Dependent variable: This is the measure that we are interested in. We are expecting it to change somehow in response to the independent variable. For our chocolate example, the dependent variable is the happiness rating given by the participant after eating the chocolate.

Confounding variable: This is a variable that we do not manipulate or measure, but one that is still likely to somehow influence our findings. For our chocolate example, the sex of the participant may be a confound (females may like chocolate more). Confounding variables can be measured, and in some advanced method of analysis, controlled for. A confounding variable is also sometimes known as a control variable or covariate.

Independent measures design: This is an experimental design where different (or independent) people take part in the different conditions of your experiment. So in the chocolate example, one group of people would have eaten white chocolate, a separate group of people would have eaten milk chocolate and a third group of people would have eaten plain chocolate. An even better example of an independent measure design would be a study looking at sex differences: the males and females have to be two independent groups of people! Independent measures design is also sometime known as between subjects design.

Repeated measures design: This is where the same set of participants repeatedly take part in each of the experimental conditions. So in the chocolate example, each person would have had all three type of chocolate and rated how happy they felt after eating each type of chocolate. Repeated measures design is also sometimes known as within subjects design.

t test: This is the statistical analysis that you would conduct if you wanted to see if there was a difference in scores between two conditions. This might be either and independent measures design or a repeated measures design.

ANOVA (analysis of variance): This is the statistical analysis that you would conduct if you wanted to see if there was a difference in scores between three or more conditions. This might be either and independent measures design or a repeated measures design.

Factorial ANOVA: In a basic ANOVA you look at differences across conditions where one variable has been manipulated, such as whether participants eat white, milk or plain chocolate. In a factorial ANOVA you can manipulate more than one independent variable. So you may also want to consider whether the amount of chocolate consumed influences happiness. You now have two independent variables: type of chocolate (white, milk or plain) and amount of chocolate (square, bar or bucket). With a factorial ANOVA you can see whether there are differences in happiness according to the type of amount of chocolate eaten, but also whether there is an interaction between the two independent variables.

ANCOVA: This is an analysis of covariance. It is just like a normal ANOVA, but it allows you to control for any variance in happiness rating that can be explained by a control variable, such as the sex of the participant.

MANOVA: This is a multiple ANOVA where more than one dependent variable can be analysed. So rather than just analysing happiness ratings, you might also want to analyse calmness and hunger ratings.