I am doing a quasi experiment with four classes at school. Two of the classes will be the control group and the other two will be the treatment group. There is one independent variable so the question is do i consider the four classes as two groups (treatment and control) and use t-test or are these four separate groups and I use ANOVA.
Thanks for your help.
Hello Mona,
Strictly speaking, if individual classes were assigned to treatment conditions, then class is your experimental unit (not individual students). This would imply a doubly nested (sometimes called hierarchical) model: You have classes within treatment condition (2 each), and students within classes (within treatments).
A simple way to analyze the data for differences due to treatment in such a design is to compute class means, and use those four means as the data points (n1 = 2 for control, and n2 = 2 for treatment). From this perspective, it won't matter whether you use independent t-test or one-way anova; you'll get the same result and conclusion (the t-test will be based on 2 df; the F-test would be based on 1,2 df). There are more elaborate analytic methods available, including hierarchical linear (aka multilevel) models, but all will have limited statistical power due to the small number of classes used.
If that sounds too fearsome because the effective sample size is only 4 cases, then the only other option is to assume that there is nothing within the individual classes that would cause students' scores to be inter-related (e.g., common effect of a really good or really poor teacher, loud environment, or students having been grouped according to some attribute that interacts with treatment condition or correlates with the dependent variable). Then, treat individual student scores as if independent cases, which is not uncommonly done in quasi-experimental studies. But do recognize, it's a mighty strong assumption. From there, good old t-test or one-way anova having 2 levels.
Good luck with your work!
Hello Mona,
I join Olena’s response and I add you should check equal variances also,after that you can use t-test. Good luck.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA