SERVE Home Page
SERVE Center for Continuous Improvement
 Overview
 Products & Publications
 Professional Development
Tools & Resources
 SBR in NCLB
 Research Studies
 
 

  

Topic Areas Emerging Issues About SERVE
SERVE > Topic Areas > Educational Research > What is SBR? > Quasi-experimental Research

 

 

What is Scientifically Based Research?

Quasi-experimental Research

We know that large-scale, random assignment experiments allow us to have a maximum level of certainty about the impact of interventions. We also know that more and more researchers are finding out how to conduct true experimental research in schools. However, there are situations in which random assignment of individual participants to experimental and control conditions may not be possible and an alternative research method is sought.

There are alternatives to experimentation that can attain a comparatively high degree of certainty while largely avoiding the barriers sometimes encountered when attempting to implement true experiments. Chief among these techniques are quasi-experimental studies. Where large-scale, randomized experiments do the best job of ruling out all unrelated explanations and leave only the experimental intervention as a viable explanation for measured outcomes, quasi-experimental methods focus on eliminating specific alternative explanations by constructing various types of specialized comparisons. Among the most important types of quasi-experimentation are:

 



  • Matching studies. Individual participants in “experimental” conditions are compared with non-participants who are at comparable or identical levels on measures of one or more relevant factors that may have an effect on outcomes. For example, students in a pilot reading program might be matched one-for-one with non-participants on IQ scores, thus ensuring that any differences in outcomes between the groups were not due to intelligence.
  • Interrupted time series. These studies employ a series of observations of a dependent variable made prior to an intervention, and then compare them with a series of observations after the intervention has been employed. This type of design is frequently used to evaluate the effects of legislative or policy changes. For example, if a state legislature enacts statewide speed limits, the effect of the legislation can be estimated by comparing the number of speeding violations for six months on either side of the point of intervention.
  • Multiple regression analysis. Multiple regression analysis, which can incorporate statistical adjustments for any number of outside factors, is often used in measuring impact. In effect, this method enables researchers to simulate a control group by statistical methods (Mohr, 1996, chapter 5).
  • Regression discontinuity studies. These studies are based on a rationale too complex to explain here but are widely regarded as among the most viable alternatives to true experiments (Shadish et al., chapter 7).