Hello.
Recently I have become more familiar with determining à prior sample sizes, but mainly with T tests.
I will need to use those procedures for a F test (repeated mesures, within-between groups) and I am struggling a little. A couple questions that could help me:
1) In the typical "non-inferiority trial" is there a value of ES that is theorically atributed in this cases?
2) Yesterday I realized that Gpower has an calculator to help determine the ES. In the category "between groups" I could determine the ES to be 0.5
Under the same exact conditions, can I use that ES in "within-between" analysis or that would be an error?
The calculato for "between groups" allow to ES to be determined by means (that is fine with), while "within-between" requires variances and is beyond my stats skills. :)
Thanks.
Hello Jorge,
With respect to your questions:
1. See this link, especially p. 153: http://hjdbulletin.org/files/archive/pdfs/431.pdf
2. For any effect of interest (within factor, between factor, interaction of the two), the issue is always, what degree of difference is clinically/practically important to detect (if, in fact, it were to exist)? This is a judgment call. The default guidelines of G*Power for "small," "medium," and "large" effect sizes come from Jacob Cohen's text, Statistical power analysis for the behavioral sciences. However, Cohen would be the first to note that they do not substitute for context-specific information and understanding of the variables. So, professional judgment is called for.
3. It is always possible to run a quick pilot study, in order to determine what sort of effect sizes might accrue in a full study. As well, this will also give you some insight as to the correlation of scores across the repeated measures dimension, which is also needed for G*Power specifications involving a repeated measures factor.
Good luck with your work!
Dear David Morse , thanks for the help. :)
I got a better idea about the non-inferiority and it statistics involved about it.
But, if we forget the idea of superior or inferiority and we would simply want to compare the means of 3 groups, given that I have an pretty good idea of the change in means in each arm (given meta-analysis and previous studies), how can I use that to estimate the ES for repeat measures, within-between factors?
Once again, in G*Power I can calculate the ES for "between factors" with that information. Example: https://gyazo.com/e426b9eefa2b5c1302fd2e5a186d9200
Can I insert the same ES for "within-between" calculation, assuming all the other parameters of the experience are the same?
In my pratical terms, my doubts arise from the fact that G*Power allows to determine ES from variance or means for "between factors", but only from variance for "within-between", while I am not so familiar with the math for the last option.
Thank you.
And well, I believe that I have found my own answer.
I did a simulation for determing ES for "within-between" and for "between factors" only, giving the same variance data in both cases and G*Power gave me the same ES for both analysis.
As such, I presume I can safely use the same ES that I have calculated for "between factors" using the data from the means in the analysis "within-between".
- Badges
- Science topic
- Similar topics
- Biological Science
- Physiology
- Physiological Processes
- Stress