One group ( Adherence group )
Three levels ( Low, medium and high adherence)
One Variable ( two sets (pre-post) of total score from 1-20, it's scale variable )
Question/ i wanna know to find differences between Low,medium,high adherence levels between pre and post-intervention
As always, this depends on your research questions. A naive approach on my part, without knowing this, particularly how you envision the use of the pre- and post-test, would be a 3X2 mixed design (adherence group between, and pre- post, within).
Hello Pshtiwan Ali. I found your description a bit confusing, but like John Morris, I think you mean you have 3 independent groups that differed in how well they adhered (low, medium, high) and two time points as a within-subjects variable (pre vs post). I also think that you are asking how to obtain pairwise comparisons for one variable within each level of the other. If that is what you are asking, here's how to do it with SPSS (which you listed as one of the topics). Assuming you are using Analyze > GLM > Repeated Measures, add one (or both) of the following EMMEANS sub-commands.
To get simple main effects of adherence at pre and post:
/EMMEANS = TABLES(adherence*time) COMPARE(adherence) ADJ(LSD)
To get simple main effects of time (pre vs post) for each adherence group:
/EMMEANS = TABLES(adherence*time) COMPARE(time) ADJ(LSD)
Feel free to change ADJ(LSD) to one of the other options. Quoting from the reference manual, "Available methods are LSD (no adjustment), BONFERRONI, or SIDAK."
https://www.ibm.com/support/knowledgecenter/en/SSLVMB_26.0.0/statistics_reference_project_ddita/spss/advanced/syn_glm_univariate_emmeans.html#syn_glm_univariate_emmeans
HTH.
First, as Bruce Weaver says, the LSD test is no adjustment of the p-values, so Safaa K. Kadhem might want to clarify this. My main point is the answer (as typical in statistics) it depends. Things like the distribution of the residuals and whether the tests are vertically scales (so 14 on the pre is in some way the same as 14 on the post) are important, but first question is whether the pre score influenced the allocation to groups. This will be important to help decide if an Ancova or the Anova (or gain score) described in other comments is more likely to be useful. Several authors have written on this. My most readable paper on this is: Deleted research item The research item mentioned here has been deleted
Hope this is useful.
Daniel Wright makes a good point about ANCOVA possibly being the way to go. But bear in mind that the groups were (apparently) not formed via random allocation. (I mention that because some authors have railed against use of ANCOVA in the absence of random allocation to groups.)
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS