sdm-help-list > Meta-analysis comparison and independence
Showing 1-4 of 4 posts
Results per page:
Jul 24, 2020  08:07 PM | john216
Meta-analysis comparison and independence
Hi all,

I did 2 separate meta-analyses to examine the neural basis of 2 cognitive functions. Now, I would like to compare these two analyses (i.e., function A > B). My questions are:

1. Should I combine the SDM tables for functions A and B and add a 'function' column (e.g., function A = 1) and use the Linear Model based on this.
2. Some studies examined both functions A and B, so the 2 meta-analyses are not independent. Is such dependence a problem for the Linear Model? If so, how should I deal with it?

Best regards,
Aug 3, 2020  03:08 PM | Lydia Fortea - Instituto de Investigaciones Biom├ędicas August Pi i Sunyer (IDIBAPS)
RE: Meta-analysis comparison and independence
Dear John

1. Indeed, you should create a column specifying the task which each paper is studying and then perform a linear model with that column to see if there is a correlation depeding on the task.

2. A good approach to deal with the dependence of the studies is to select one funcion for each study (maybe you can perform several analysis to check if the result change depending of which function you select). On the other hand, if there are just few studies that used both function; you can add both studies as independent by assuming  that the error would be small as there are few studies.

Kind regards
Aug 4, 2020  08:08 AM | john216
RE: Meta-analysis comparison and independence
Dear Lydia,

Thanks for your reply! I have 21 studies included, and 11 studies examined both functions. If I have to choose only one contrast for each study, the sample size for each meta-analysis would become much smaller. Is it possible to conduct a hierarchical meta-analysis in SDM to maintain the sample size for each function?

Also, I don't know how to test the effect of the selection on the results with this amount of overlapped studies. Should I select contrasts randomly and do the meta-analyses multiple times?

Aug 5, 2020  04:08 PM | Lydia Fortea - Instituto de Investigaciones Biom├ędicas August Pi i Sunyer (IDIBAPS)
RE: Meta-analysis comparison and independence
Deat John, 

Unfortunately SDM doesn't have the option hierarchical meta-analysis. In your case, when performing the linear model you need to choose one function for each of the 11 papers that studied both, in a way that number of papers that studied each function are balanced (you have an example in If you want to perform two meta-analysis indepedently you can include all.

The performance of several meta-analysis by considering different factors for each paper can be a good solution to check if the results remains the same or they vary too much depending on the seleected papers.

Kind regards,