Hi All,
I'm now currently working on a data set and I'm unsure of whether mixed ANOVA is the appropriate method to analyse my results. We want to see whether our intervention is useful or not. Here's the research design:
A 6-session workshop (intervention) is conducted to parents who's child do or do not have mental disorders. Pre-tests were done before the intervention started and a post-test at the end of the intervention. I'm hoping to see whether the intervention works for both groups of parents (those with or without kids with mental disorders).
Using mixed-ANOVA, the within-subjects factor is time, the between-subjects factor is with or without kids with mental disorder, the dependent variables are parental stress level, self-efficacy, psy well-being etc.. I'm wondering is using mixed-ANOVA the correct way to analyse my results? As all the averaged mean scores of DVs of the group without mental disorder is higher than those with mental disorders in the pre-test level. Is this going to cause any impact to the interpretation of results? If yes, what can I do to get meaningful results?
Also, if mixed-ANOVA is not the way to analyse, what should I use instead? (I'm using SPSS to analyse my data).
Thank you so much.
Hi Jo Ch ,
It does sound like you are on the right track with this analysis, as long as
your outcome variables are continuous variables (.." parental stress level, self-efficacy, psy well-being etc. .."). (Make sure you screen their distributions well - ANOVA is a parametric (normal distribution) test.
Your design is a 2 (between-subjects: group) x 2 (within subjects: time). The design is a classic "split-plot" design - the analysis will take a difference score (pre-post) for all subjects and then perform a between-groups test on the differences, so the fact that the groups are different to begin with is not as big a problem (people are serving as their own controls in a sense, because only their change is analyzed). A bigger problem is not that the between-group means are statistically different, but if one group is at a ceiling', i.e. they do not have any room to grow or change.
If you have another variable (not an outcome variable) that you could use a co-variate you might want to explore that in the ANCOVA framework. A good discussion of covariate use in non-randomized trials like yours is Chapman & Miller (2001) (http://www.bwgriffin.com/gsu/courses/edur8132/notes/Miller_Chapman_ANCOVA_Abuses.pdf).
As a final note - even though it is a common misnomer, the presence of both a between-subjects factor and a within-subjects factor does not constitute a 'mixed' design, mixed refers to the presence of fixed and random factors in a design. Your is just a factorial ANOVA.
Hope that is helpful,
It sounds correct. In case you want to compare your groups on a number of different but RELATED dependent variables you have to use MANOVA. I attach one of our intervention studies:
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA