I am just curios which variables to match. If one variable is likely to be a risk factor, like sex, do we need to match our cases and controls by sex in case-control study?
Thank you in advance.
Dear Mulugeta,
Matching in research is used when there is a well known confounding variable. It is known that these kinds of variables affect the risk factor and the outcome at the same time. so it inculcates bias in your study and it affects the validity of your study.
The first criteria whether a variable should be matched among cases and controls is depends whether that variable is a confounding variable or not. This is confirmed by literature. There are two types of matching: 1. individual matching and 2. frequency matching.
The way you analyze your data is different for the first type which is conditional logistic regression. for the second type is is not different.
For more you can read Modern epidemiology!
Good luck!!
Dear Tegene,
Thank you so much and your thoughts are much appreciated.
One more question: What if the risk factors to be tested are three or more? Let's say we are planing to explore: living with HIV, having Psychiatric problems, sex and taking NSAIDS as risk factors for a definite outcome. How do we handle these variables during matching? Thanks.
I don't know how many people you plan to enroll in your study but three risk factors would indicate a competing risk model. You want to do a search on propensity score matching. Also, what statistical software will you be using?
Dear Mulugeta,
The number of variables to be included as a risk factor doesn't have a problem if you have sufficient number of samples. But if you are going to make the listed variables as a matching variables it is possible but you may not get enough samples and you don't measure the magnitude of the risks for your matching variables.
As the number of matching variables increases you are approaching to make the cases and controls similar for their exposed risk factors. the matching variable should a well known confounding factor or risk factor.
Good Luck!!
Thank you Joseph and Tegene. Your thoughts are much appreciated. I believe I will get 30 cases and 60 controls; but I don't have any idea on what statistical tool to use. I have to ask my statistician. Thank you again.
before the answered I need to know your intention to use matching why you need ? then I will continue
Thank You Yemaneh. We just need to control the variables of potential confounders.
- Badges
- Science topic