Covariates can be described as characteristics (excluding the actual treatment). You could collect data about characteristics before running an experiment to determine how treatment affects different groups and populations. You could also use this data to adjust for any covariate.
Study results may be affected by covariates. You might run an experiment to determine how corn plants cope with drought. While the “treatment” is the actual “treatment”, it is not the only factor that influences how plants perform. Size is also a factor that can affect tolerance levels so you could include plant size as a covariate.
Covariates can be independent variables (i.e. It can be a variable unrelated to the direct interest or it could be a variable confusing it. The accuracy of your results can be increased by adding a covariate. In ANCOVA The independent variables are categorical variables. You might, for example, look at the effects of different treatments on depression. independent variables The type of treatment received. The dependent variable is a countable variable. It’s like a self-scoring depression score. There is however a lot of unexplained variation within the group. This is because people do not enter the experiment with the exact same level of depression. You need to be aware of your initial depression level.
The covariate in this context is always:
- Observed/measured (as opposed to a manipulable variable).
- A control variable
- A continuous variable.
Another example, from Penn State: Let’s suppose you compare the salaries of men to determine who makes more. You need to keep in mind that people tend to make more money the further they have been out of college. This is called a covariate.