Dear everyone,
I implemented a quasi experiment pre and post test design. A new teaching intervention in class.
as other researchers did, I used pre test as covariance and run ANCOVA to see effect of intervention on treatment group compared with control group.
However, my data violated the assumption of homogeneity of slope.
I already seek some possible solution but it is quite complicated to me.
+ Johnson-neyman technique
+ Multilevel analysis
Have you ever encountered this case? What technique you applied?
Thank you so much.
When the slopes are unequal it means that there is an interaction between the IV and the covariate.
Dear Thanh Huynh Quan,
Multilevel analysis would be a good possibility. However, you need to consider your sample size. It depends on whether you have both conditions realized in all classes, or not. In the former case, the estimation of parameters starts getting reliable from 10 classes upwards. In the latter case, you'll need more classes. I'd choose a random slope model to model the unequal slopes. If you have predictors on the class level that might account for the slopes, even better. I've written a paper about such problems, unfortunately only in German (Schoppek, 2015).
Dear Thanh Huynh Quan ,
You should run a multiple linear regression instead.
Peter Samuels thank you sir,
as i know, ANCOVA is actually multiple linear regression.
- Badges
- Science method
- Similar topics
- Teaching