Tools for Those Who Summarize the Evidence Base
Earth scientists project temporal trends for global temperature far into the future, even centuries. Such estimates are valuable tools for civic engineers and city planners to avoid potential catastrophes.
But here's my question: Have any of you seen the practice applied in a meta-analysis? Can you think one that tried to project how a particular phenomenon, say, hate crimes, might appear if observed levels of homophobia reached zero? Or, perhaps a projection about how much longer people would live if inequality were reduced a certain amount? It would seem such a practice could be quite useful under some circumstances.
Let me know, please!
We now have a manuscript in press at Research Synthesis Methods describing just how you would make such projections and giving strategies for doing so in two platforms (Stata and R). The moving constant strategy, as we call it, is useful for both confidence intervals around an estimated effect size and for confidence bands across a moderator dimension.
We imply in the paper how conventions across the sciences for display of meta-regression results vary widely. This paper may help those who have never "moved the constant" to do so and benefit from clearer displays about their results. It may also help to avoid artificially bifurcating continuous variables (just to present results for separate levels). Instead, you can move the constant and generate whatever estimates you may need.
All, this article has now appeared at RSM.