(1) Much of Dr. Shadish's current work pertains to experimental and quasi-experimental design. Some of this work is theoretical, including the revision of a classic book on experimental and quasi-experimental design. But he works extensively on several related empirical projects. Those projects have the shared theme of describing and understanding differences between experiments that randomly assign subjects to conditions (randomized experiments) and those that do not (nonequivalent control group designs). Most of this work has used meta-analysis, a set of quantitative techniques for reviewing and integrating literatures. Initially, the effort was to examine whether or not the two kinds of designs yield the same effect size. More recently--and more importantly--the effort has been to understand the circumstances under which quasi-experiments do a better and worse job at approximating the effect sizes yielded by randomized experiments. We have done this work in such diverse topics as SAT coaching, marital and family psychotherapy, presurgical psychoeducational interventions to improve postsurgical outcome, juvenile drug use prevention, ability grouping of children in classrooms, occupational therapy, and alcoholism treatment.  Dr. Shadish's laboratory has recently developed a laboratory analogue paradigm for investigating this question, as well. In this paradigm, participants are randomly assigned to be in either a randomized or a nonrandomized experiment (in which they then self-select conditions). This paradigm allows unbiased estimation of the difference between randomized and nonrandomized experimental results, exploration of different design methodologies to reduce this difference, and application of statistical models for the same purpose.

(2) Dr. Shadish also works extensively in meta-analysis. Much of this interest is in studies of the methodology of meta-analysis itself. One such line of work has been the application of mediational modeling to meta-analytic data--instead of just describing outcomes, can we say something about the processes that led to differential outcomes? How could meta-analysis contribute to this? A similar line of work concerns the computation of effect size. Over the years a host of methods for such computations have taken place, but without much systematic criticism of how well they approximate each other. This line of work has used computer simulation and actual data sets to examine this question, and he has recently published ES, a computer program for estimating the standardized mean difference statistic from 40+ different kinds of data (http://www.assess.com/ES.html). Finally, some of Dr. Shadish's work on meta-analysis is application, doing meta-analyses in various areas to study questions of substantive interest. Most of this has been on psychotherapy research, including one review of the effects of marital and family psychotherapy, and another on whether the results of psychotherapy research generalize to the kinds of settings and conditions under which psychotherapy is usually conducted.

(3) Dr. Shadish has written extensively about program evaluation theory. Some of this work overlaps with the preceding work on methodology because experimental designs and meta-analyses are often used in program evaluations. But program evaluation theory incorporates much more than just this. In addition to other kinds of methodology, program evaluation includes the study of epistemology and ontology, of value theory, of how information gets used in decision making, and of how social interventions change an improve both naturally and with deliberate attempts to change them. All this comes together in recommendations about how program evaluations should actually be done, both in general and in specific cases. Consequently Dr. Shadish spends a good deal of time consulting to others about how evaluations can best be done.