I want to check the effective role of auditors in improving the financial reporting quality through meta-analysis and literature is available for both internal auditors and external auditors in this issue.
My objective is to explore which type of auditor (internal auditor vs external auditor) is more effective in promoting financial reporting quality through meta-analysis.
In a meta-analysis if will conduct subgroup analysis or divide all studies into two different parts (studies with internal auditor and external auditor) and iI will run separate meta-analysis on both types of studies or run sub-group analysis of these two types of studies.
Can anyone please suggest me what is most the appropriate test for the comparison of the output of two meta-analytic outputs or comparison of different subgroups in meta-analysis?
thanks in anticipation
Hi, in my opinion, conducting a mixed-effects model (or meta-regression) with the moderator represented as one or several dummy variable, the most straightforward procedure. You get a B-coeffizient which reflects the mean difference in the effect size between the chosen category and the reference category as in any other dummy regression. This model can be applied for both fixed effects or random effects meta-analyses as well as three-level-meta-analyses in which you have multiple effect sizes per study.
HTH
Holger
It is possible to run meta-ANOVAs. Lipsey and Wilson have a macro for how you can run this in SPSS. http://mason.gmu.edu/~dwilsonb/ma.html
Hi Bilal,
Based on the Schmidt and Hunter's (2015) approach to meta-analysis, in order to test the significance of a moderator using sub-group comparison, you just need to check the confidence intervals for each sub-group and ensure whether they overlap with each other. If they overlap, it indicates that there is no significant difference between the two sub-groups (i.e. two levels of a moderator). On the other hand, if they don't then you conclude that the moderator is significant. A more sophisticated approach to determine the significance of a moderator using the confidence intervals is given in attached articles if you like to consider for your project.
Hope this helps.
You have two options:
1. You can test for interaction of two meta-analytic estimates which is a Z test to see the difference between two independent estimates from subgroup analysis or two separate meta-analyses. This is fully explained in "Altman DG, Bland JM: Interaction revisited: The difference between two estimates. BMJ 326: 219, 2003".
The problem with the above approach is the low power of the test. There is also risk of wrong results due to nature of data (confounding).
2. The other option is to run meta-regression. It is more robust but to perform the meta-regression you need a minimum of 7 to 10 studies. You can find more detail of the assumptions and interpretation of the results in "Thompson SG, Higgins JPT. How should meta-regression analyses be undertaken and interpreted?Statistics in Medicine 21:1559, 2002".
Hope this helps.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis