Skip

How can I do a meta-analysis of just two papers?
January 17, 2013 10:37 AM   Subscribe

I'm working on my first ever scientific paper: a meta-analysis of data from other studies. I've just finished my search for relevant papers. My paper is to cover two related (but different) topics (determined by my advisor). For Topic A, I have a fair number of papers (nearly 20) from which to obtain data. For Topic B, though, I only have two! Is it even possible to do a meta-analysis of two papers? If so how would I go about it? I'm asking y'all instead of my advisor because he is unavailable this week and next, and I want to go ahead and get going, at least by reviewing other meta-analyses so I can get a sense of what I need to do with the papers I've got.
posted by ocherdraco to Science & Nature (9 answers total) 4 users marked this as a favorite
 
IANAResearchScientist, but I've definitely seen meta-analyses (usually of medical literature) that analyzed small numbers of papers and concluded that the evidence to date was insufficient to support any firm conclusions, that more research was needed in X/Y/Z directions, etc.

Also, are you sure you've really exhausted the resources on Topic B? Any closely-related research questions you could draw in there? Have you mined the lit reviews and bibliographies of the two papers you've got, to see if they cite any other studies that might have slipped through your keyword searches? Have you checked out the websites of the authors of the two B-papers, to see if they're currently doing any additional research on the same topics?

Lastly, if the weird disparity in the quantity of research on A vs. B is interesting at all, that'd also be something to comment on in your paper.
posted by Bardolph at 11:07 AM on January 17, 2013 [1 favorite]


Did your advisor ask you to do a meta-analysis? My understanding of meta-analysis papers is that they require some tricky statistics and a lot of knowledge in the field. I'm not sure if it is something suitable for a first paper.
posted by Lord Force Crater at 11:10 AM on January 17, 2013


Yep, that's what I've been asked to do.
posted by ocherdraco at 11:15 AM on January 17, 2013


Some good advice here at Stackexchange.
posted by cnanderson at 11:23 AM on January 17, 2013


I think it is going to be really hard to say without knowing what the topic is and what you are trying to analyze about it. I've never been part of a meta-analysis so I'm probably not your guy in any case, but I imagine that only having a sample size of N=2 for Topic B will put serious constraints on the types of statistical analysis you can do, yes.
posted by Scientist at 11:24 AM on January 17, 2013


I think I am going to proceed with Topic A as planned and ask my advisor about Topic B when he is available.

I think it is going to be really hard to say without knowing what the topic is and what you are trying to analyze about it.

I know, and I'm sorry about that. I think I've gotten the information I need to know in order to move forward for the next week or so, which is the main thing I needed. Thanks, all.
posted by ocherdraco at 11:29 AM on January 17, 2013


I think I am going to proceed with Topic A as planned and ask my advisor about Topic B when he is available.

That sounds like a good plan. I've never done a meta-analysis, but have read many, and have never seen one examining only 2 papers - that seems like a waste of time, really. I suppose it could be field-specific, though, so your advisor is the best one to ask.
posted by randomnity at 12:06 PM on January 17, 2013


I just (ten minutes ago) gave a library rearch class to a bunch of chemistry students. We talked about multiple ways to make sure our searches are exhaustive. Consult the reference librarian for your subject on your campus. She/he may be able to help you find more articles. She/he will also be able to show meta-analysis or review articles in your field.
posted by mareli at 12:11 PM on January 17, 2013 [1 favorite]


Every answer here is almost exactly wrong. It's incredibly common in medicine to combine just two studies in a meta-analysis. Cochrane is full of them. Two underpowered negative RCTs, but with a similar homogenous result can be combined to find a significant effect.
Nothing about the actual meta-analysis needs n>=3, but if you only have 2 studies, then the quality of your systematic review is extremely important relative to the quantitative synthesis, since the result may depend on bring sure you have all literature, and being sure it's appropriate to combine what you have.
posted by roofus at 12:48 PM on January 17, 2013 [2 favorites]


« Older I love coffee, and I have it e...   |  After a more than 10 year long... Newer »
This thread is closed to new comments.


Post