0
$\begingroup$

Background: Group A the control group uses a website 1. Group B the treatment group uses website B. The expected result is that Group B will have a higher understanding than group A. Hypothesis: The treatment group will show increased levels of understanding. A 1(Strongly Agree) - 7(Strongly disagree) likert survey was used, to measure understanding after the groups used either website 1 or 2.

I know using a two sample t-test will allow me to determine whether a significant difference was present. But how do I determine the direction of the difference through using SPSS? Still new to this. Thanks for the help.

$\endgroup$
  • 3
    $\begingroup$ In the tables SPSS outputs for T-TEST the sample means are right there to tell you which group is higher or lower. $\endgroup$ – Andy W Jun 2 '14 at 15:13
0
$\begingroup$

If there is a significant p-value for the t-test (for example, p < 0.05), the confidence intervals (CI) will NOT overlap. For example, let's say Group A has an average understanding of 3.5 (95% CI: 2.5-3.75), and Group B has an average understanding of 5.5 (95% CI: 4.5-6.5), we can conclude that we are 95% confident that the mean understanding of subjects from Group B is higher than the mean understanding of subjects from Group A.

On the other hand, if there is not a statistically significant p-test value for the t-test (for example, p > 0.05), the CIs will overlap. For example, let's say Group A has an average understanding of 3.5 (95% CI: 2.5-3.75), and Group B has an average understanding of 4.5 (95% CI: 3.5-6) -- In other words, there is overlap in the confidence intervals! -- Thus we do NOT have evidence to conclude that the mean understanding of subjects from Group B is different from the mean understanding of subjects from Group A.

I hope this helps!

$\endgroup$
  • 4
    $\begingroup$ Sorry Matt, but I think this is incorrect: Non-overlapping CIs can still be "significant" (i.e. $p\leq0.05$). It's true that if the CIs don't overlap, a significance test will yield a significant p-value. But the opposite is not true. See here and here for example. $\endgroup$ – COOLSerdash May 25 '18 at 21:41

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.