I'm looking at an intervention's efficacy on three different age groups (so not randomly allocated between conditions) and need to analyse the following:
- are improvements in each of the groups separately significant. Within participants analysis
- comparing the level of improvement post-treatment between the three age groups. Between participants
How would I go about this? Any advice appreciated as I'm getting bogged down in reading about ANOVAs and T-tests and the examples do not seem to reflect this design.
Also not sure if a control is necessary as this is motor skills in neurodevelopmental disorders, which don't improve on their own without therapy.
Many thanks
Hello Alison Kaye
I think it is very easy to use the One-way ANCOVA plan (if you do not have a control group). See the link below
https://statistics.laerd.com/spss-tutorials/ancova-using-spss-statistics.php
If you have a control group, use the Two-way ANCOVA plan.
If you are interested, see also the following links.
GraphPad Statistics (2020). Two-way ANCOVA analysis with SPSS. Statistical tutorials and software guides. Retrieved January, 12, 2022, fromhttps://graphpad.ir/general-linear-models-univariate-two-way-ancova-spss/
GraphPad Statistics (2021). Before-after data, analysis of covariance or repeated measurements?. Statistical tutorials and software guides. Retrieved January, 12, 2022, from https://graphpad.ir/repeated-measures-or-ancova
Assuming that your improvement outcome is numeric, here is what I see as the simplest and easiest to interpret:
1) First, do a paired t-test to test for each subgroup separately. This will test whether the pre vs. post scores (or other numeric outcome) are significantly different, or in other words it tests whether the average change in score for a subject is different from zero.
2) To compare the three groups, you can use ANOVA on the whole population with the dependent variable being the change in score for the subject and the indpendent variable being the age group. You can also compare any two groups with a 2-sample t-test of the same variable.
You should be fine without a control if there is no reason you would have a placebo effect.
No, you use Chi square tests to analyse your type of data. The above two replies by Abolfazl Ghoodjani and Tina D. Hunter contain incorrect advice. You do not require controls with ANOVA (there's no such thing as a control anyway since they comprise actual treatments). And you never do separate t tests on a big data files because they ignore interactions and will be different from ANOVAs. I have specialist training in experimental design btw :)
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA