Evaluation Methods:
True Experimental Design
Also referred to as "randomized experimental
design,” this is the best real world approximation
to the ideal experimental design, in which
the evaluator randomly assigns subjects to
treatment (program) and control groups.
Experimental design is the strongest design choice when there is interest in establishing a cause-effect relationship. Experimental designs for evaluation prioritize the impartiality, accuracy, objectivity, and validity of the information generated. The intent is to make causal and generalizable statements about a population or impact on a population by a program or initiative.
Non-Experimental Design
Non-experimental studies use purposeful sampling
techniques to get “information rich” cases.
Non-experimental evaluation designs include:
case studies, data collection and reporting
for accountability, participatory approaches,
and mixed method studies.
Quasi-Experimental Design
While causal inference generated based on true
experimental designs are usually very strong,
such designs can be difficult to administer
and implement. Quasi experimental designs
take a more practical approach. Control groups
can still be used, but these have to be assigned
through some non-random process, or the researcher
cannot control which group will get the treatment.
Alternatively, one can also examine program
beneficiaries before and after exposure to
the intervention.
Like the experimental designs, quasi-experimental designs for evaluation prioritize the impartiality, accuracy, objectivity, and validity of the information generated; they also look to make causal and generalizable statements about a population or impact on a population by a program or initiative. Examples of quasi-experimental designs include: comparison group pre-test/post-test design, time series and multiple time series designs, and non-equivalent control group. counterbalanced designs.