STATISTICAL MANAGEMENT AGAINST EXPERIMENTAL WORK PLANNING
Keywords:
Active experiment plan, Optimal model, Statistical analysis, Management of obtained experimental resultsAbstract
The mathematical theories concerning the plan of experimentation are relatively new. These were born in
the 70s, and their rapid development and application is being encouraged by the intensive use of computers. Early
statistical methods (regression analysis, statistical correlation, etc.), which were mainly used for mathematical
modeling, supported the processing of results from the "passive experiment". During this, the mathematical methods
of processing these results (collected data), are put into service in the last stage of an important process towards the
formation of the mathematical model for the object under consideration. The notion of experiment plan which is
used here, means laying out experiments based on a scheme previously designed in the form of the respective
matrix. This means that specific mathematical methods must be used at all stages of the experiment. This can be
decomposed in advance into the following notions: during the analysis of the relevant notes; before and during the
experiment planning; during the processing of the results arising from the experiment; in the research strategy
selection phase; in the final stage for the correct interpretation of data and results.
The classical approach to conducting an active experiment, when only one factor (variable) changed, while the
others remained at the initial level, is practically not used in contemporary industrial facilities, because this method
does not take into account the interactions of factors. Also, in this case a large number of experiments would have to
be performed, but the reliability of the results would be at a low level. The cost of experimental research, according
to current knowledge, is proportional to the square of the number of experts participating in them, while their
efficiency is proportional to the square root of the total number of experiments. This means that experimentation is
required to be optimized, because it does not make sense to make a very large number of experiments, the value of
which would exceed the value of the positive effects that will result from their results. By optimizing the experiment
it is meant to choose the most suitable mathematical model, but which requires a minimum number of experiments.
It is rightly assumed that the active plan of experiments represents a possible optimal plan
References
Avetisov, S. E., Novikov, I. A., Bubnova, I. A., Antonov, A. A., & Siplivyi, V. I. (2010). Determination of corneal elasticity coefficient using the ORA database. Journal of refractive surgery, 26(7), 520-524.
Bellman, E.R., & Dreyfus, E.S. (2013). Applied Dynamic Programming, Princeton University Press, Princeton, New Jersey
Churchman, C.W., Ackoff , R.I., Arnoff, E.L. (2017). Introduction to Operations Research, John Wiilley and Sons, New York
Harvey, W.M. (2018). Principles of Operations Research, Prentice-Hall, New Jersey
Kocherlakota, N. R. (1990). Disentangling the coefficient of relative risk aversion from the elasticity of intertemporal substitution: An irrelevance result. The Journal of Finance, 45(1), 175-190.
Manson, N. J. (2006). Is operations research really research?. Orion, 22(2), 155-180.
Stoller, D.S. (2017). Operations Research, Process and Strategy, University of California Press, Los Angeles
Vajda, S. (2013). Mathematical Programming, Addison – Wesley Publishing Company, Inc. London
