Many researchers use correlations in their studies, but it is not without risk. This lesson explores some of the issues that researchers may experience if they use correlations.
One of the best statistical tests out there, in my opinion, is the correlation.
Correlation is a mutual relationship between two variables. This means that when one variable goes up, the other will respond by increasing (known as a positive correlation) or decreasing (known as a negative correlation). Easy-to-understand variables include:
- The more one studies, the higher their grades are.
- The more one eats, the less hungry one feels.
But in the social sciences, there is rarely a perfect correlation. A perfect correlation means simultaneous and equivalent changes are seen when a variable is altered. For instance, if you were doing an experiment on studying and you saw a perfect correlation between hours spent studying and grades on tests, then you would look at your data again.Behaviors and real life are so complicated that you rarely see a perfect correlation.
There will be something running interference. You, as a researcher, want to try and reduce the amount of interference.There are other limits to correlation, and the researcher who uses correlation techniques needs to be aware of some of the limits of correlations. And the real scary part is that they’re not always as obvious as the perfect correlation issue.
Correlation does not equal causation.
Experiments are typically designed to demonstrate that variable A causes something in variable B. That’s exactly what our previous examples of studying and eating are all about. However, those examples are incredibly simplistic.
Here is a question that has a little more complexity: Does violent television cause people to be more violent?The question starts to flip back and forth the more you think about it. Does violent TV make people more violent, or do violent people watch more violent TV?A problem with correlation is that the variables you are interested in are merely interacting with each other. They are not necessarily causing one another. So whenever you are using a correlation, it is inaccurate to say variable A causes variable B.
All you can say with a correlation is that variable A interacts with variable B.
Correlations do not indicate direction of interaction.A correlation does not tell you how variable A and variable B interact.
All it tells you is that they do interact. Here is another example: As drug use increases, interpersonal relationship problems increase.This means that as drug use increases, relationship problems also increase.
Or is it that as relationship problems increase, there is also an increase in drug use? All that we can tell you from our correlation is that the two are connected.Our example aside, we can make educated guesses as to the directionality of a correlation, but the correlation itself does not tell us which variable causes what we’re observing. Our educated guess is based on our understanding of a subject, the experiment, and what we can interpret from the correlation.
Correlations may be measuring a third, unknown variable.This is one of the most common concerns when using a correlational design. What happens when you’re not actually measuring the two variables you think you’re measuring? Here is the famous example: Murder rates and ice cream sales are positively correlated.
As murder rates increase, so do ice cream sales. Apparently, after a good murder, the killers like to go out and buy ice cream. Make sense; murder takes a lot of work.
What is actually happening is that a third variable is being measured. After some work, it became evident that murder rates and ice cream sales were both correlated with the temperature. As temperature increases, so do tempers and ice cream sales. The original correlation of ice cream and murder was being influenced by an unknown, unseen variable.There is no simple way to tell if there is a third variable messing up your data.
You have to take a good, hard look at the variables you are studying and what they are saying. The math helps here, but it is more of a logic problem instead of a math one.
Correlation is a mutual relationship between two variables.
A perfect correlation, which is a simultaneous and equivalent change seen when a variable is altered, almost never happens in psychological research. Correlation also has several other limits, which a researcher must be aware of. This includes:
- Correlation does not equal causation.
- Correlations do not indicate direction of interaction.
- Correlations may be measuring a third, unknown variable.
When this lesson is done, you should be able to identify the three main limitations to correlation.