Experimental Design in Ecology (Part III)

or
"Quantitative synthesis and meta-analysis: 

linking data with models and moving beyond P-values"

Where:  University of Pittsburgh's Pymatuning Laboratory of Ecology

When:  July 28 - Aug 1; M-F 8-5pm

Instructor:  Craig W. Osenberg (personal web page)

Course Information:

• Course description
• Assignments to be completed before the first day of class (this is very important)
• Schedule and readings

The course will be structured like a workshop, with a combination of lectures and discussions, and considerable time to work on group projects.  The central theme of the course will be meta-analysis -- the quantitative synthesis of data from independent studies that test a similar hypothesis.  As a result, we will revisit some basic premises that we've all learned in our statistics and ecology courses, including the role of null hypothesis testing and the meaning and interpretation of P-values.  As an alternative, we will discuss the derivation and estimation of effect sizes, and more importantly, we will struggle with the very difficult issue of defining a suitable measure of effect size.  Defining "effect size" is largely a conceptual problem and one that can't be solved with statistical expertise alone (e.g., how do you quantify the "effect of competition" or the "relative importance of top-down control" or "seed limitation" or "phenotypic plasticity" or any other phenomenon of interest to ecologists?).  To highlight these issues each student will participate in a meta-analysis project.  To jump-start the process, there will be some work required before the first day of class....

Prior to the start of class on Monday, July 28th, please:

1. Read Gurevitch and Hedges (1999), Osenberg et al. (1999), and Johnson (1999).  
2. Think about a meta-analysis project that you would like to do.  On Tuesday morning (the second day of class), you will need to present a 2-page proposal for a meta-analysis project.  This proposal should include a compelling conceptual motivation (why is the question important and how can meta-analysis be used to illuminate the problem), as well as detail about the availability of papers that can be used in the meta-analysis.  You will turn in a copy of your proposal to the instructor at the start of class and give an oral presentation to the the class.
3. Try to have at least the beginnings of a list (full citations) of ~20 papers (certainly, no fewer than 10 and no more than 30) that contain data that can be used in your proposed meta-analysis.  Most of you will probably already have such a list derived from your graduate research interests.
4. Try to have copies (hardcopies or pdfs) of these papers (or at least most of them) with you at Pymatuning (this will save us considerable class time).