I would like to perform a meta-analysis using an index of feeding preference for insects (quotient between number of insects fed from a particular host divided by the relative abundance of such host). Every study will include only the number of insect, the percentage of (part of) them fed on that particular host and the relative abundance (%) of that host. It will not include a mean (only a ratio or quotient) neither a standard error. I would like to know if it is possible to apply a meta-analysis without this information (not because it is not available per se, but because it is not applicable to this type of data). I know that if I use only porcentage I can perform a binary meta-analysis (I have 2 dichotomous variables) but I do not how to do a meta-analysis for this type of index. If any could help me (or recommend me some bibliography) I would be very grateful.
Thanks in advance, Kevin.
I don't know if I understand your desired index. I suggest that you familiarize yourself with risk ratios and odds ratios and doing meta-analyses on them. (See the attached PDF.) It is also possible to calculate an "event rate" and do a meta-analysis on that. I do not have a book reference for event rate, only a screen shot showing its calculation.
Thank Gordon L Warren for your response. A binary meta-analysis could be apply for my data only if I use percentages (and it fits in your explanation above). But according to what I have read about meta-analysis to include a study it has to report means and standard deviation (if I am not wrong). However what I want to evaluate is the behaviour of an index which is a combination of two variable: proportion of insects fed on a particular animal divided by the proportion of this animal in nature. It is not just a porcentage, is an index which range between 0 and infinity. So I think I cannot apply a binary meta-analysis. In fact, I do not if I can apply a meta-analysis at all (that is my question) due to that in a paper it in not reported as a mean (is just a single number) neither is applicable a standard deviation. Thank for the answer anyway.
No, meta-analyses can also be done using binary data (e.g., odds ratios, risk ratios). The "RR_&_OR_calculations.pdf" file I provided is from a well-known meta-analysis textbook.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis