0
$\begingroup$

I have a research of 30 patients (one sample) which performed cognitive tests in the following states:

  1. Baseline
  2. at 3 months with no intervention
  3. at 6 months after intervention

These are the same patients who underwent all 3 states.

The cognitive measures are continuous parameters and I am using SPSS. I am wondering how to analyze them:

  1. Paired t test of 1-2 compared to 2-3
  2. Repeated measures with 1 as covariate ? is this "legal"?
  3. splitting as 2 groups of 1-2 and 2-3 so each has a pre score and a post score and performing a ANOCVA with the pre score as covariate (i.e. for 1-2 group the covariate values will be 1, and for 2-3 group the covriate values will be 2)

I realize that this is an atypical cohort, but any help will be greatly appreciated.

$\endgroup$
  • $\begingroup$ What specific question do you want to investigate? $\endgroup$ – Michael M Apr 29 '17 at 17:24
  • $\begingroup$ Thanks for the reply. I want to investigate if the cognitive functions improved after the intervention or not. The first and second points were made for test-retest validation and removing a possible "learning" effect. $\endgroup$ – XPeriment Apr 29 '17 at 21:55
  • $\begingroup$ help anyone? ? ? $\endgroup$ – XPeriment May 2 '17 at 18:30
0
$\begingroup$

Your design is clearly worse than a pre/post design with randomization into control and test and clearly better than an intervention only pre/post design.

In case of no (time-dependent) confounders I would favour suggestion 1) as it is simple and easy to explain to others. I guess you are aware that it assumes constant placebo effect over time. To avoid relying on normality assumptions, maybe go for a permutation t test. This should be possible even in SPSS.

Option 2) could be interesting in case you want to take into account (time-dependent) confounders.

$\endgroup$

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.