1
$\begingroup$

How does one compare means between two groups that contain repeated measures. For instance:

Group 1 = has mutation A

Group 2 = has mutation B

Dependent variable = blood pressure (continuous variable)

Group 1 has 140 data points and 52 participants. Group B has 120 data points and 30 participants. Some participants have five measurements of blood pressure, others have four, and yet others can have only one. This is not a "repeated measures" analysis because we are not studying the effect of time or diet or anything like this on blood pressure. Some participants simply have more measurements than others.

Approach 1. Take one random measurement per participant.

Approach 2. Take the first measurement from each participant.

Approach 3. Take the average of all measurements from each participant.

Approach 4. Take all the data points and run a t-test ignoring the problem.

Approach 5. ?Something better? Please say that this is possible.

For bonus marks, could someone tell me how to implement this in SPSS?

Thanks for your time.

$\endgroup$
1
$\begingroup$

We do collect repeated measures to see the change in response between subjects as well as variation within an individual over time. A repeated measure design is powerful, as it controls for all potential sources of variability. So the problem you described is definitely a repeated measures analysis problem.

For definite reason i would go with approach 5. You are to think of a linear mixed models for the data. To compare means with repeated measures, you can use "One way repeated measures anova" which is equivalent to univariate paired t-test in this regard.

In SPSS

  • go to Analyze-->Generalized Linear Models-->Repeated Measures
  • define your response and covariates
  • define the factor you are interested to compare the means
$\endgroup$
  • 1
    $\begingroup$ For a repeated measures generalized linear model, how would you define the within subjects factor in such a situation? For example, the blood pressure measurements could be 30 days apart, 60 days apart, 90 days apart, 105 days apart, or any other number. Doesn't a within subjects factor need to be consistent between subjects? $\endgroup$ – Angry_at_Linux Jul 7 '14 at 20:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.