For an outcome (primary endpoint), I am using Continuous and Generic Inverse Variance Data Type to perform meta-analysis. As the data representation is not uniform; studies represented data either as the end of study mean +/- SE (Continuous) or mean difference +/- SE (Generic Inverse Variance Data).
Is it possible to summarize the outcome in the single Forest plot?
Can Fisher or Edgington methods solve this problem? if yes, please help me to perform meta-analysis with these methods.
Thank you ...
Please elaborate ...
How to perform meta-analysis with Fisher or Edgington methods?
When you stated "mean" that implied one group data, whereas "mean differences" that implied two groups' or pre-post comparison data. Can you try to sort out these design or grouping stuations in your data management? If so, then I would suggest to code and sort the data further before pooling data for forest plots.
Depending on the numbers of studies per designs (one group or comparisons) and for the interest of publications, in general, the more studies at the higher level of designs (hierarchical level for scientific evidence) for a given scientific field, the better chances for you to publish your meta-analysis in higher impact journals. If the number of studies is less than desired, you can identify common themes for different levels of designs or grouping for pooled analyses (I see systematic reviews rather than meta-analysis commonly used in paper title for this approach).
I think it is not possible because de differences betwen two means, could be appears in order of 1 o 2 points and the mean probably appears at 20 or something, then you differences will appears in the bottom of the drawn and probably it couldn´t works for you.
I did some statistics and finally got a solution to pool all data (mean+/-SE and mean difference) in the same Forest plot (Data Type - Generic Inverse Variance).
Few studies expressed data only in mean difference form and others either in mean+/-SE (also change from baseline) or mean-difference. Therefore, I used Generic Inverse Variance (with Totals and Sub-totals) Data Type to pool all data. The studies which reported data only in mean+/- SE form, I calculated mean difference by meta-analysing the means, and confidence intervals (95%) were used to calculate standard error (SE=(upper limit - lower limit)/3.92).
PS: @ Pamela Shiao, thank you for your answers. It gave me the idea to pool all data in the same Forest plot.
Apart from Fisher's or Edgington's methods, Stouffer’s method is also used for Meta-analysis methods based on P -Values.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis