My experiment has 2 factors with 2 levels. It will be conducted in a lab. there are 7 dependent measures (ordinal in nature). If I do the experiment once to measure all the 7 outcomes (includes 3 mediator variables)--- I believe 120 is an acceptable sample size.. but if I decide to split the study into 2, measuring 5 outcome variables in 1 experiment, then replicating the same experiment to measure the remaining 5 outcome variables, will I have to double my sample size to120 +120??
Hello Stuti,
All other things equal, the more DVs (dependent variables) in your model, the larger the sample size will need to be to achieve a specific power level. So, if you've already determined that N = 120 is a reasonable lower bound for an analysis with 7 DVs, then that should be more than sufficient for a study with fewer (either 2 or 5) DVs.
However, the question I would have is whether using the same cases in both studies would somehow introduce a possible confounding effect in the second study (due to exposure to the conditions and measures from the first study). If that were possible, then you would need two independent samples, each of sufficient size for your research aims.
Good luck with your work.
Thank you David Morse but I am contemplating going ahead with one run of the experiment..its not a repeat measures design anywys..so its minimum 30 unique respondents per cell. 30*4 = 120.
It really makes no sense doing the experiment twice. Rather I will aim to collect more than 120 responses so I can support my results (with the argument that I have maintained a sample more than 120 therefore providing better robustness than recommended standards) What do you say?
Another question now bothers me.. For verifying the manipulation check and reliability validity of construct measures.. what would be the size for pilot study?
Stuti Kumar , I guess the first question is how you determined 120 was appropriate. You presumably input the minimum effect size to detect plus other parameters into a power analysis, and it gave 30 per cell, which would be low for many fields of study, but could be fine for your context. Anyway, say how you did that and the power analysis can be expanded. When you combine things one additional thing you have to do is tell the computer what counts as a "success", for example, would you need both "studies" to yield a certain effect size or p value, or say at least 6 of the 7 measures?
I guess the first question is for you to say what the input and output were for you power analysis.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size