For testing the effectiveness of group intervention, I would like to conduct ANCOVA. I have one experimental and two comparison interventions. However, my data is not normally distributed. Also, I have a small sample size. So, I was wondering if there is an option to run nonparametric ANCOVA in SPSS?
We need more info. First if you want to run ANCOVA you must have covariates. Please tell us about those. The same with your depoendent variable. In particular what is it.and how was it measured. You say your data set is not normally distributed. Describe what you mean and how you know about the distributions? please tell the sample sizes, how the groups were selected and what do they consist of. With this info we should be able to at least begin to help you. Best, David Booth
Dear Prof. Booth,
Thank you very much. I am testing the effectiveness of a psychological intervention as a Randomised Controlled Trial. I have three groups with very small sample sizes. First one has 17, the second one has 11 and the third one has 10 participants. I have pre and post-test scores (self-report instruments). I would like to use pre-test scores as a covariate since groups were not matched based on pre scores.
Regarding normality - Although skewness and kurtosis values are in the range of + / - 2, normal distribution value for Kolmogorov-Smirnov or Shapiro-Wilk indicates non-normal distribution.
I hope this is sufficient.
Kind regards,
Başak
Dear Prof. Conroy,
I have one active control group where I also do an intervention and one wait-list control group. My hypothesis is that my experimental condition would result in a greater decrease from pre test to post-test compared to the control groups.
Kind regards
If you are familiar with R, you can use sm.ancova package to access Non-parametric ANCOVA test.
Wadie Abu Dahoud thank you very much. However, I am not familiar with R. Therefore, I asked whether it is possible to do it in SPSS.
One approach is to run a partial regression (excluding the primary factor of interest) and then perform a non-parametric analysis of the residuals. This is described in Koch et al (1998) Article Issues for covariance analysis of dichotomous and ordered ca...
However, see Lesaffre and Senn (2003) Article A note on non-parametric ANCOVA for covariate adjustment in ...
for a necessary correction to this approach.
See also Akritas (1996) Article On the Use of Nonparametric Regression Techniques for Fittin...
and Thas et al (2012)
Article Probabilistic Index Models
Your data is nonlinear with mean, variance, skewness & kurtoses of the distribution, that may be the first four terms of infinite Taylor series expansion representation, so why not to try Bayesian parametric framework of maximum likelihood estimation? If yes you may follow https://www.researchgate.net/project/Statistical-Learning-on-manifolds-with-its-applications-in-computer-vision?_sg=vUPagzea3Dj3honJa0MieXfihrvbXTS6_IUmo40skPQlCgTNNJknKpgVKQN6SHLw9xa7HWjCS1R9aXR0bULAwLIJUnvpGQwEed87
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running