I've searched and there are no online examples I can find for conducting a similar M-A using CMA. I have means, for example, minutes/day for different occupational groups. These are not means from intervention/control or mean differences, just means minutes/day of individuals within a study. I want to obtain 1) the overall mean for all individuals, 2) the means within each separate occupation group (comparison) and 3) whether the means of the comparison groups differ.
To obtain the total effect size should I use option 1 or 2 from below? The values are different depending on which I use.
I am doing:
Study A nurses A
Study B nurses A
Study C nurses A
Effect size nurses A
Study D postal delivery B
Study E postal delivery B
Study F postal delivery B
Effect size postal delivery B
Overall outcome A and B together
While the other way is:
Study A (regardless of occupational group)
Study B
Study C
Study D
Study E
Study F
Total effect size
Should I also run outcome A in a separate M-A to outcome B to ensure that the outcome is specific to the occupation group? For example:
M-A #1: Nurses only:
Study A nurses A Study B nurses A Study C nurses A Effect size nurses A M-A#2: Post-delivery?
OR should I run all occupation groups together to get a total mean and separate comparison means (like the first example I showed). If I do this, the effect estimates per occupation group are affected by the studies included in the full M-A. I assume then it would be better to do a separate MA for each occupation?
Hey Syephanie,
Unfortunately I am using R for such analysis and not familiar with CAM, however, I would like to record my thinking. You have three questions:
1) the overall mean for all individuals, 2) the means within each separate occupation group (comparison) and 3) whether the means of the comparison groups differ.
The first one is easy and you can do it in any software. The second and third question can be answered in Meta-Analysis. For question 2, you can run MA separately for both groups and you will find Effect-Size (mean etc.) for each group and do comparison then. For question 3 you can run for the whole data to get the over all Effect-Size.
I have a paper in which I did MA, however, it is about the mean differences and related to hydrology. You can see it and may be help you out:
Article Improved Hydrological Projections and Reservoir Management i...
Best of luck
Thank you Firdos. I appreciate the reply. Apologies 1) should be overall mean across all studies. I believe that is still required which I think is the answer you provided for #3.
Hi Stephanie,
I think you are looking to conduct a meta-analysis on 'raw means' rather than the typical mean differences than meta-analyses typically study. A great resource could be ESCI software (check https://thenewstatistics.com/itns/esci/) which is quite easy to use for this type of data (also with two groups. It comes with some good documentation of similar types of examples (like milliseconds response time to something).
Otherwise, as long as you compute the standard error of the mean, you can also use some Excel workbooks we developed (check www.meta-essentials.com and http://onlinelibrary.wiley.com/doi/10.1002/jrsm.1260/full) - consider using Workbook 1 with mean as ES and compute the standard error based on the SD of the mean.
If you are just interested in the overall mean across studies, I would recommend to first run the basic analysis (irrespective of group) and secondly, (if there is sufficient heterogeneity to explain) split out the studies over occupational group using a subgroup analysis (possible in both ESCI and Meta-Essentials). The subgroup analysis (using a test for significant differences or ANOVA) will also tell you the answer to your third question.
I hope this helps. Let me know.
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running