Correct Analysis Method for 1:1 and 1:4 Matched Studies with Continuous Outcomes?

I have several questions about analyzing 1:1 and 1:M matched studies that have a continuous outcome variables.

1.To determine difference in 2 means (continuous outcome) for a 1:1 matched case-control study the proper test to use is either a t-test or a Wilcoxon Ranks Test. What tests can you use if you want to use regression on a 1:1 matched design? Can you use a basic linear regression model (proc reg in SAS) if all the assumptions are met? If a nonparametric model is needed, what do you recommend?

2.When trying to find the difference between 2 means (continuous outcome) in a 1:4 matched case-control study, what is the proper statistical test to use? (I know it is incorrect to use the t-test and Wilcoxon Ranks Test). Secondly, if you want to apply a regression model is there a special kind to use for 1:4 matched studies with a continuous outcome? What type do you use if all assumptions are met? If assumptions are not met is there a nonparametric version?

For a 1:1 case control study that has a dichotomous outcome it is proper to use either the McNemar’s test, a Mantel-Haenzel matched-pairs analysis. For a 1:4 matched case-control study with a categorical outcome it is correct to use the PROC PHREG (conditional logistic regression) procedure. These tests cannot be used for a 1:4 matched design with a continuous outcome variable.


Less is more. Stay pure. Stay poor.
To get you moving along, I believe you would want Proc GLM with Strata in your statement for linear/non-normal linear and Proc Logistic for categorical outcome. Not sure PHREG will do much for you (depends on data and goals). Which ever regression procedure you use for 1:1, I would also use for the 1:M.
Thank you, I will look into Proc GLM with Strata for matched studies with conditional outcomes.

From what read there are different stats for 1:1 vs. 1:M designs, but I'm having a hard time tracking down what exactly they are.

From my hunt for the correct stats to use, I came to understand that proc logistic can't be used for a MATCHED design with a categorical outcome because of the conditional vs. unconditional logic. The PROC PHREG is a SAS code for conditional logistic regression. Proc logistic is for unmatched (unconditional) designs with a categorical outcome.
Last edited:


Less is more. Stay pure. Stay poor.
Example of Proc Logistic with matched data:

proc logistic data=dataset_name order=data;
	class variable1;
    strata subject;
    model dependent_variable_name = variable_1 variable_2/ rsq;