Hello,
I made an experimental study with a pretest-posttest control group. I treated the pretest as a covariance variable. However, since my data were not normally distributed, I performed a nonparametric ANCOVA (Quades method). But I don't know how to report the result. What would you recommend as a sample article?
ANCOVA is just a particular type of ordinary least squares (OLS) model, so it is the errors that are assumed to be normally distributed (with a mean of 0 and variance = sigma2). Did you examine normality of the residuals from your model, or the dependent variable itself? And how did you do the assessment? Thanks for clarifying.
Dear Weawer,
I know the ANCOVA vs. nonparametric ANCOVA assumptions. There are different approaches adopted in non-parametric ANCOVA analysis. e.g. . Quade Method, Puri & Sen Method,McSweeny & Porter Method.I adopted the Quade Method. The Quade method is based on testing the equality of the residuals among groups obtained from the linear regression of the ranked response variable and the ranked covariate. This method is distribution-free and the test statistic shows asymptotically central F-distribution. The Quade method is based on testing the equality of the residuals among groups obtained from the linear regression of the ranked response variable and the ranked covariate. My assumption is based on that. My dependent variable is not normally distributed. I also analyzed this distribution according to Z-scores. I'm looking for a sample article that takes these assumptions into account. Thank you for your help.
Hello Menşure Alkış Küçükaydın. You wrote:
My dependent variable is not normally distributed.
This suggests (to me) that you have not examined the residuals from an ordinary ANCOVA model. IMO, you need to do that before deciding that ANCOVA is unsuitable.
Understood.
Thank you for your professional support and interest.
- Badges
- Science topic
- Similar topics
- Correction