A perspective: After administering the tools before and after the intervention, you will analyse the data separately. The results for pre- and post- intervention will then be placed side by side for Paired t test.

Example, for SF-36, the results for the 8 domains and the 2 summary components (PCS and MCS) of both periods (pre and post) will be placed side by side, and the difference (lets call it K) will be computed by simply running Post minus Pre column.

A mean of 0 means no change.

IF your sample is random, and your observations are truly dependent (i.e. pre and post done on same individuals), and the differences or K (i.e. Post minus Pre) are normally distributed (usually determined by running simple histogram) or your total sample size is big enough (>=30), then you can test mean of 'score difference' (i.e. whether the score after intervention is different from 0 or not).

In SPSS: Analyse--> Compare Means --> Paired t test --> Post - Pre

The output will report Mean (SD) of Pre value, Mean (SD) of Post value, Mean of score difference (95%CI) and P value.

A significant +ve P value (

Thank you so much!

In addition to the excellent recommendation from Mohammed, you could run repeated measures ANOVA with scores added in at time 1 and time 2. This does the same thing as mentioned, however it gives you the added benefit of being able to add a covariate to the model if you are concerned about any extraneous variables influencing the relationship.

For example, if you do not have a control group and you show an improvement in social connectedness and physical health from pre to post intervention, you can't be certain that the intervention is responsible for the increase in these scores. Depression is associated with Loneliness, but a different construct all together and could explain the change in scores. Say though, you also measured participants' depressive symptoms, you could add this into the model as a covariate to rule out the alternative explanation that people's depression subsided and that's why they reported more social connections and better physical health.

Thank you for the added methodology! If instead of using the same group I were to have a control group to randomize the experiment, would I be able to use paired t testing and ANOVA as well?

Hi Stephanie,

I am a little bit confused by your follow up question. Do you mean that you have a control group and a treatment group, and they are randomised between participants? Or does everyone do both the control and the treatment, but it is randomised as to which they receive first (i.e., control first then treatment [AB], vs treatment first and control second[BA])?

If it's the first, you can compare with a mixed model ANOVA. If you're using SPSS, you would start it off like a repeated measures ANOVA (i.e., analyse-> general linear model -> repeated measures) and then add the between participant condition as a factor in the model.

If it's the second, you will need to use a similar model but this time including condition sequence as the between participant factor (i.e., AB, BA) to see whether there are order effects. You can then use paired-sample t-tests to compare the cell means.

I hope this helps and I haven't misunderstood your question.

Thank you for your response! I meant the first scenario you mentioned. Both the control and treatment groups would receive a pre- and post-test evaluation. However, only the treatment group, of course, would receive the intervention.

When the sample is same paired t test need to be used to see the effect of intervention