2
$\begingroup$

I have about 50 pre & post-surveys for an intervention. I am limited to using excel and was told to only calculate the change scores.

For example:

Change Score    # Participants

-2              0
-1              6
 0              15
 1              16
 2              4

            41 total participants

(16 [change score of 1]+ 4 [change score of 2]) --> 20/41 (49%) showed improvement following the intervention.

For some reason this is not sitting right with me; if I had access to SPSS I would've assessed using ANOVA.

Any tips? Opinions?

$\endgroup$
  • $\begingroup$ Hi, Hannah. What exactly is not sitting right with you about the distribution of the change scores? Would you expect fewer participants to have experienced improvement? If so, bear in mind that you only have a sample of 50, and it is possible that there is selection bias favoring subjects who experienced positive change. $\endgroup$ – Brash Equilibrium Jun 13 '18 at 19:39
  • 1
    $\begingroup$ (And if you could elaborate on what is not sitting right with you, and on what you want in an answer, that would be very helpful.) $\endgroup$ – Brash Equilibrium Jun 13 '18 at 19:40
1
$\begingroup$

49% showed improvement, 37% showed no improvement, and 15% regressed. There is some evidence here of a positive treatment effect.

To statistically analyze the data you want to do a paired t-test. You can search for paired t-test in Excel for plenty of resources. Also, since the data don't look continuous you may want to perform a non-parametric test, such as Wilcoxin ranked test. You can compute this test manually in Excel without too much trouble. Here is one resource a quick search found: http://www.real-statistics.com/non-parametric-tests/wilcoxon-signed-ranks-test/

| cite | improve this answer | |
$\endgroup$

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.