I am trying to calculate the sample size with G-Power software for a study, but I have the following problem: I have not come across any other study with similar groups and blood parameter, at least similar to the study in my head. Therefore, I am having trouble calculating the effect size based on means. I am thinking of a cross-sectional study with 4-5 groups, in this case, what would be the right way to follow for sample size?
I can help you in this ,Select proper statistical Analysis Factorial Design and ANOVA model to get sample size in G power tools
Hello Caner,
Yes, often the most challenging part of a priori power analysis is identifying a defensible target effect size (ES).
If there's no prior research to guide you, for the specific variable of interest, then you can:
1. Ask domain experts their opinion as to what would comprise a clinically important ES. You could pose the task as: (a) some degree of difference in mean values across conditions; (b) some degree of explained variance due to condition differences; (c) some standardized mean difference (e.g., if treatments outperformed control/customary condition by, say, 1/4 SD) as examples.
2. Run a small pilot study.
If you already have in mind some pattern of mean differences that, to you, would signify a noteworthy difference, then you'd also have to declare some common group SD or variance. Once that's done, it's pretty simple to convert to a Cohen's f (which is the default ES in g*power for anova) from within the program (using the "determine" option).
Good luck with your work.
- Badges
- Science topic
- Similar topics
- Correction