As part of my thesis I'll be conducting a single-sample pre-post intervention analysis, in which my only DV is ordinal (7-point Likert scale). In my research proposal I said that I would use paired samples t-test for this analysis, to compare means in the sample before and after.
In a meeting I received feedback that this data analysis wouldn't be sophisticated enough. I didn't get more info than that, and I'm very much a novice when it comes to statistical analyses. I'm hoping to get some advice here on what could constitute more sophisticated analyses in this pre-post intervention study.
The only idea I've come up with is to test whether certain participants (eg based on gender, age, etc) are more likely to show a significant difference post-intervention. I'm not sure of the test I'd use to determine this, though. I'm also unsure whether this is 'sophisticated'.
I sincerely appreciate any help here.
A paired sample t-test would be appropriate if the dependent variable is continuous and normally distributed. However, since your dependent variable is ordinal, the paired sample t-test is not suitable. For ordinal data, you would typically use a non-parametric test like the Wilcoxon signed-rank test for paired samples.
J. Fraser, there are several more sophisticated statistical methods you could consider for analyzing pre-post data, especially when dealing with ordinal data such as a 7-point Likert scale. One approach is the Wilcoxon signed-rank test, which is a non-parametric alternative to the paired samples t-test. It does not assume normality and is suitable for comparing paired differences in ordinal data. Another option is ordinal logistic regression, which can model the relationship between your ordinal dependent variable and one or more independent variables, providing insights into how factors like gender or age influence outcomes on your Likert scale. Mixed-effects models, also known as hierarchical linear models or multilevel models, are useful for handling repeated measures and can account for both fixed effects (such as the intervention) and random effects (such as individual differences among participants). Generalized estimating equations (GEE) offer another method for analyzing repeated measures data by accounting for the correlation between repeated measurements and handling various types of dependent variables, including ordinal ones. Latent growth curve modeling, a type of structural equation modeling, allows you to analyze the trajectory of change over time, modeling both the initial status and rate of change in your dependent variable while accounting for individual differences. Additionally, analysis of covariance (ANCOVA) can be used to compare pre- and post-intervention scores while controlling for covariates such as age and gender, helping to account for initial differences that might affect the outcome.
To implement these sophisticated analyses, you might consider using OriginPro software. OriginPro offers comprehensive tools for advanced statistical analyses, including those mentioned above. It is user-friendly and can help you perform sophisticated analyses with ease, addressing the feedback you received and enhancing the overall rigor of your thesis.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis