| Module 3: Interpreting Data
5. Association between two variables
5.1 Revision and introduction
Studies which are not framed in terms of hypotheses may be aimed at establishing whether there is association (relationship) between two or more variables for the same group of individuals or units. We may even want to know if it is possible to say that one variable causes changes in another variable. As you learned in Module 1, demonstrating causation in observational studies can be extremely difficult.
Association between two variables means there is a relationship or connection between them. [We will also use the term "co-relation" or "correlation" as well as association - these are interchangeable terms.]
In the introduction to this module we mentioned the study which examined the effect of pollution on the birth weight of babies in Sydney. The amount of pollution was strongly associated (correlated) with the varying birth weights of babies. One expert said that it was a causal relationship.
When you are reading a report about the association between two or more variables, you must ask the following questions:
- What units or participants do the data describe?
- What are the variables and how are these measured? (This will indicate to you the quality of the variables used to measure a characteristic.)
- Are all the variables quantitative or are some categorical? (This should influence the way that the data are presented:
- Categorical variables will require bar charts, pie charts or two-way tables when two variables are being examined.
- Quantitative variables will require line graphs when there is only one variable being examined, and a scatter plot when there are two variables being examined.
- If there is a relationship between two variables, how strong is it?
- Does one variable cause changes in the other variable? (For example, with variables such as the level of alcohol consumption and problem solving ability or level of alcohol consumption and the ability to drive a motor vehicle, a causal relationship is often argued.)
- Which variable explains the variation in the other variable - i.e., which variable is the response variable?
- Is any association due to a third intermediate or confounding variable?
Module 1 covered the use of experiments to produce data allowing researchers to claim that one variable, called the explanatory variable, caused a change in another variable, called a response variable. In experiments, efforts are directed towards controlling all other variables so that claims of causal relationships between the explanatory variable and the response variable can be substantiated. For example, does a cholesterol-lowering margarine really lower cholesterol or is there some other mechanism involved? In observational studies the causality of any relationship between two variables is more difficult to substantiate, as you will learn in the following sections.
|  |
