I want to do a systematic review and meta-analysis on certain topic. when I looked into available articles it is too less. at least how many should I have to do so?
Thanks!
It‘s better to have an experienced statistician who can give advice on topic selection or conduct group discussions.
Hello Yosef,
I interpret your "less" as "there is a low number of primary studies". If this is correct, these are my experiences:
Scientific disciplines vary with regard to the number of primary studies that give rise to a meta-analysis (ranging from medicine to ecology, for instance). Two aspects which are essential are
a) the heterogeneity of effect sizes
b) the importance of coming up with an average estimate fast (e.g., in medical treatment).
I had once a submission to a journal where we did a meta-analysis in which almost all primary papers had found a somewhat positive correlation. Hence, the reviewer responded with "we all know there's a relationship, so why do we need a meta-analysis?". This cured me from doing meta-analyses when there is now heterogeneity.
Statistically, you can estimate a mean in a fixed-effects meta-analysis with two studies. A random effects meta-analysis needs more (around 7-10) to estimate the true variance among studies.
As stated above, the main question is as in any other paper: Do the field needs a summary?
HTH
Holger
Yosef Eshetie
"A paragraph from the Book "Advances in Meta-Analysis" written by Pigott, Terri.
Another common question is: How many studies do I need to conduct a meta-analysis? Though my colleagues and I have often answered “two” (Valentine et al. 2010), the more complete answer lies in understanding the power of the statistical tests in meta-analysis. I take the approach in this book that power of tests in meta-analysis like power of any statistical test needs to be computed a priori, using assumptions about the size of an important effect in a given context, and the typical sample sizes used in a given field. Again, deep substantive knowledge of a research literature is critical for a reviewer in order to make reasonable assumptions about parameters needed for power.
Good Luck ....
http://www.springer.com/gp/book/9781461422778
https://www.researchgate.net/post/Is_there_any_minimum_number_for_studies_that_should_be_included_in_meta-analysis
The meta-analysis component will depend slightly upon the statistica modeling you will employ (fixed vs random vs mixed), but many recommend a minimum of 5 independent (i.e. not from the same study/cohort) effect sizes to satisfy model parameters and provide meaningful insights. A systematic review can have less studies and be okay, so long as it provides a thorough breakdown of why there are so few, discusses the major strengths/limitations of the exisiting, and reasons why more evidence is needed. It will likely be harder to publish however (if that is your goal) if there are very few studies included as limited studies is a similar issue to limited sample size in original research, but will depend upon your research topic and field overall. Hope this helps.
There is no perfect number, but you might need to adapt the search terms to receive additional hits. Maybe check out the following guidelines: Article A New Paradigm for Systematic Literature Reviews in Supply C...
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis