Minimax design - construction, effectiveness comparisons, and practical applications


Project leader


Funding source

Swedish Research Council - Vetenskapsrådet (VR)


Project Details

Start date: 01/01/2010
End date: 31/12/2012
Funding: 2670000 SEK


Description

As experimental methods are used more frequently in the social sciences they need to be developed further. Proper planning of experiments is important because it will increase the certainty in the conclusions. Properly planned experiments are also cost effective since they are more reliable given a certain number of observations. Alternatively, fewer observations can be used for a given reliability. A method with great potential for experimental design is the minimax design. It has so far not been used to its full extent because of some mathematical difficutlies. A newly developed algorithm that can handle these problems has provided preliminary good results. The purpose of the project is to study minimax design. Three partial purposes can be outlined: (i) development of the algorithm and evaluation of its convergence properties, (ii) exploring the efficiency of the design in comparison to other methods and (iii) applying the design to practical problems. For example, bidding experiments are used to measure the value of environmental goods that are otherwise hard to price. Questions in ability tests, such as College entrance tests, can be tested in experiments. A researcher can use the algorithm to decide on the number of difficulty levels, which levels to use, as well as the number of questions to use for each level in order for the test to be optimal.

Last updated on 2017-31-03 at 12:58