Tools for Those Who Summarize the Evidence Base
I am relatively new to the world of 'meta-analysis' and have some questions about how to undertake a meta-analysis and what kind of data I can include.
I want to investigate the influence of variables pertaining to island biogeography (size of islands, distance to the mainland, etc) and life history variables (type of development, PLD, etc) to the genetic diversity of aquatic invertebrates on oceanic islands.
My initial understanding was that I could collect data from the literature, i.e. find the level of genetic diversity (i.e. heterozygosity) of a species (per island) from one paper, and then collect additional information about the species and the island to eventually build up a database with several variables I hypothesize would influence the level of genetic diversity (dependent variable) on islands, and then carry out statistical analyses on this database (I was under the impression at the beginning of all this that I would just conduct traditional statistical analyses, ANOVA and regression, etc, but since beginning to delve into the literature on meta-analysis, I now understand that this is not the correct thing to do).
But, since reading through the literature on meta-analyses, I have come across the 'effect size'. I understand what this is (I think), but I also don't think genetic diversity (the variable I am interested in) is a measure of an effect per se. I don't have the difference between two means, I don't have a correlation coefficient, nor do I have an odds ratio (the three main types of effect size statistics I have been reading about).
So, my main issue is this 'effect size'. If I am not using an effect size statistic, as I am just collecting the overall level of genetic diversity per species per island (reported in journal articles as heterozygosity or haplotype diversity), am I not carrying out a meta-analysis?? Or am I somehow able to convert this to an effect size??
Any suggestions or guidance or comments on my problem would be greatly apprepriated,
Thanks for your time
It sounds like you could do a meta-analysis of these data, yes, but it would not be one with a two-variable effect size (e.g., OR, r, or d), it would be a one-variable effect size: heterozygosity. It would be analogous to doing meta-analysis with the arithmetic mean or of the proportion. To date there have not been many of these meta-analyses published, so it is not easy to recommend a model. I discuss the concept in my 2008 paper, applied to the level of social dominance that samples exhibit. Lipsey and Wilson (2001) discuss the concept in their 2001 book as well. In order to apply the idea, you would need the mean level of heterozygosity for each island, the standard deviation of heterozygosity, and the number of samples. If you can get this information from the source reports you indicate (or estimate it), then it would seem feasible to meta-analyze.
Thank you for the reply, I will look at both your paper and the book you suggested. I haven't found any examples to date that use a one-variable effect size, but as I said, I'm in the early stages of reading through the literature that is out there (which is a lot, especially in the field of ecology, even though it is a new-ish concept adopted in the field, it certainly has made a large contribution to the field so far), so hopefully I will find something.
Thanks again for replying and helping me out.