Hi,
I want to do a meta analysis from GWAS and publications. For the GWAS studies, I only have the beta value, p value and total number of samples while for the publications I have sufficient data that could be used to calculate effect size.
Is there a way to compute effect size from only the beta, pval and N?
If not, is there any way to calculate beta using the mean, SEM and N of cases and controls from publications?
Thanks in advance!
Hello Dhanya,
The answer is, it depends. If you're happy evaluating regression coefficients as part correlations (adjusting for the other IVs in a model) for your target ES, then yes. Just treat it ("beta") as an adjusted correlation. However, if one study invokes a dozen IVs, and another only five, and a third uses one, and a fourth uses 29, then are the coefficients comparable if one of the IVs is common across all four studies? No, not necessarily. As well, one presumes that the DV is common, or functionally identical, in each study.
A correlation can certainly be an ES metric, and the sample size gives information as to how tight the CI around that value would be (e.g., precision of the estimate).
If this isn't at all what you were after, perhaps you could elaborate your query.
Good luck with your work.
You can use the method implemented in METAL (https://genome.sph.umich.edu/wiki/METAL_Documentation) which uses uses p-value and direction of effect (from beta), weighted according to sample size.
Just to muddy the water GWAS IS very complicated and There's multiple uses of beta in statistics. Perhaps we should have a look at this method before answering this question. Fora start see the attached search. Best wishes David Booth
Another useful search is given here. Note variable selection is sometimes used here for regression prediction equations. Several studies have used poor choices for such methods..This can impact final.results and hence a meta analysis. A very brief introduction and an interesting paper is referenced in the second attachment about the variable selection problem in GWAS. Best wishes David Booth
- Badges
- Science topic
- Similar topics
- Analytical Chemistry
- Column