Proposal #9416177 In many fields, researchers have traditionally found it economical to do research by creating and testing a few, well though-out experiments in the expectation of a high successful rate. Nature takes a different approach -creating many experiments and then screening for the few successful ones (random mutation and survival of the fittest). The cost-effectiveness of each experimental strategy depends on the relative cost of designing experiments versus "running" them and analyzing the results. Today, the relative costs of experimental steps associated with the design of new products and processes is being radically affected in many fields due to the use of new or greatly improved experimental methods such as simulation, mass screening and genetic algorithm techniques. Often, the effect of these techniques is to make "nature's" experimental approach increasingly cost-effective. This, in turn, is changing the way experimentation is being (or should be) done in many fields. This research examines the economics of the new forms of (often, computer assisted) experimentation in a number of fields, and to compare them with the economics of more traditional experimental methods. The findings of this research will provide academics with new insight into the economics of experimentation in the design of new products and processes. The findings will also enable developers of new products and processes and governmental decision makers to make better judgments with respect to the cost-effectiveness of experimentation strategies they may use during the development of new products and processes. For purposes of analysis, the execution of an experiment as involving a four-step cycle: (1) one conceives of or designs an experiment; (2) one builds the (physical or virtual) apparatus needed to conduct that experiment; (3) one runs the experiment; (4) one analyzes the result. Each of these four steps has a cost. The absolute and relative magnitude of these costs can affect exp erimental strategies. Thus, experimenters might logically strive to replace expensive experimental methods with cheaper ones. And, if step (1) - design of an experiment- is much cheaper than step (3) - running that experiment- one might logically expect an experimenter to invest more in step 4 -design- if by doing so he or she can reduce the cost or frequency of required experimental runs. Recently-developed experimental techniques such as simulation, mass screening and genetic algorithm techniques can radically affect the absolute and relative cost of the four steps involved in experimentation. 15 pairs of experiments (30 cases total) are studied. Each pair will contain one experiment performed in a manner that is traditional for the filed studied, and one experiment of a very similar type performed using a novel experimental technique such as those mentioned just above. Analyses of cost and benefits of these experimental pairs are followed by the creation and testing of hypotheses as to which techniques will be most effective in experimental conditions characterized by known levels of key variables such as the cost of generating materials or designs to be used as experimental inputs.
