Dears,
I did a framed field game exiperment with 4 different treatments i(ncluding the the control) whereby I have 45 people/players in each treatment. Can you share me your opinion on whether 45 is a good sample size to make a robust comparison (significant test, regression) between treatments?
Hello Elias,
The answer depends on several factors. First, how large or small of a difference among groups is worth detecting (or paying attention to)? If you're only interested in large differences, chances are very good that your sample size of 45 per treatment condition will be adequate. On the other hand, if even small differences are clinically/practically/educationally important, then 45/group may be insufficient.
The way to be sure is to run a power analysis. The G*Power program (freely available at: http://www.gpower.hhu.de/) can assist with this process. Since group differences appears to be the issue of interest, I'd suggest trying "F test" (for test family) and "ANOVA: Fixed effects, omnibus, one-way" for the specific test type. You'll need to declare a desired effect size (using Cohen's f: this carries the information about how small/large an effect is worth detecting), desired alpha level (type I error risk), desired amount of statistical power (likelihood of correctly rejecting a false null hypothesis), and number of groups (for your study, 4).
Good luck with your work!
You should interpolate expected effect size from previous lit to determine sample size vis-a-vis power params.
Hope this helps,
Matt
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size