Stage 8. Processing of the expert information, its analysis and presentation for DMP

 

SUBJECTS: Organizer, Prognosist, Expert.

GOALS AND PROBLEMS OF THE STAGE (The goals should be realized in the shown sequence.)

1. To process the expert appraisals [Prognosist].

2. To analyze and comment them [Prognosist, Expert].

3. To submit the formed generalized conclusion of experts to DMP in a structurized form with a compulsory formulation of conclusions [Organizer, Prognosist, Expert]. For a realization of the latter problem it is necessary to solve a totality of sub-problems mentioned below in the last part of this stage.

 

DESCRIPTION OF THE STAGE:

The generalized conclusion of experts must be formed within the framework of this stage through processing of all expert conclusions and analysis.

The first level of the stage. First of all, the processing of both quantitative and qualitative results of the interrogation. In this case both a discrete and a statistical approaches are used.

There are comparatively few works where the problems of the processing and analysis of a single expert appraisal were mentioned. In a majority of works there is an analysis and processing of the information obtained from a group of experts.

More often the procedure of the statistical processing of expert data includes three sub-problems, how this was done in [31]:

· an analysis of every expert’s appraisals;

· a calculation of the objects group appraisals;

· an estimation of the group appraisals trustworthiness.

The first sub-task - a check-up of a formal correctness of indicating appraisals by experts. For example, in case when an expert gives his appraisal in a form of a grouping it is necessary to check that sub-multitudes into which a multitude is being divided with a classification method do have blank intersections. In case when experts give appraisals obtained with the multiple comparisons method or with the ranging it is necessary to check up a correctness of the binding ranges calculating.

A definition of the group appraisal is one of the problems of the statistical processing.

In that case when the opinions of experts differ, and this fact happens almost always, it is necessary to introduce such a concept as "averaged" opinion in one or another way and to solve the second sub-problem. A selection of means’ types depends on the kind of expert appraisal and, consequently, on the scale in which an appraisal is given.

A choice of the best alternative is one of the most important problem arising during the expert information processing. The start of research works on this problem is related to the end of the XVIII-th century (French scientists Condorce and J.Sh. de-Borda). At present time this trend is intensively progressing.

Among other problems of the analysis and processing of the expert information it is necessary to note the following three:

¾ a coordination control of individual expert appraisals values, a revelation of some sharply deviating values, either two or more probable sub-groups of coordinated values of appraisals in the entire totality;

¾ an obtainment of the value of a generalized meaning of a coordinated group of the individual expert appraisals and evaluation of their coordination (dispersion);

¾ a development of non-manipulated procedures with the help of the game theory for N-players: it is the so called an active expertise.

If not few interesting works [23,40,91,108,155] are devoted to a solution of two firstly mentioned problems then the situation is slightly more complicated in relation to the last problem. The works of A. Gibbard [150], H. Moulin [162], V.N. Bourkov [13,14] and some other works are practically exhausting the list of research works on this problem.

In the case, when the experts give a totality of their appraisals of some or other indices directly or indirectly related to each other then it is necessary to coordinate these appraisals (in the sense of a correlation with each other) in a such way that regularities of relations between indices were reflected in the expert appraisals correlation of these indices.

So, if to consider the economic indices then their estimations, according to the opinion of I. Fisher [148], should correspond to a test of the "coordination" and "complementarity". For example, the rates of production growth are related to the rates of labour productivity growth, and the latter - to the rates of the productive investments growth. There are no doubts that the coordination of the expert estimations of various economic indices allows to improve the quality of the economic prognosis in a sufficient way.

In the concluding and the most important part of this stage let’s consider a forming of generalized conclusion of experts for DMP in the form of a structured document - the prognosis.

This part of the stage is extremely complicated and has a creative character. It is solved in an interesting way in the work of V. Platt [70] on similar problems and, therefore, we take his way of raising the problem as the basis.

Now, when prognosis was prepared as an information document, Organizer and Experts should solve the following problems in corresponding to the final task set in the sixth stage:

1. To find the method of stating the expert information.

2. To define the structure of the document and its size.

3. To make a decision whether to expound the conclusions in the form of categorical assertions or to give some comments for them.

4. To make a decision about what tables, illustrations etc. should be included into main document and in an enclosure to it.

5. How to make entire document to be logical and well-considered.

6. How to show DMP even in the first paragraphs of the prognosis that our material is worthy of reading, learning and having constantly in mind.

7. How to "synthesize an objective a priori information and information obtained as a result of expertise with the purpose of reducing it into a form convenient for making decisions" [9].

8. How to draw up a prognosis in a such way that it would stir up a confidence of DMP.

Therefore, as far as possible, we have to work up a short, convincing and comprehensible prognosis for DMP", and the conclusions should ensure a realization of prognostic task set in the sixth stage.

 

NOTES:

1. This extremely important stage is described in numerous works among which it is possible to mention the following Russian research works: V.N. Bourkov [13,14,15], A.A. Dorofeyuk [28,29], E.A. Eltarenko and E.K. Kroupinova [31], B.G. Litvak [49], B.G. Mirkin [59], Yu.N. Tyurin, A.P. Vasilevich, P.F. Andrukovich [92,93,94], D.S. Schmerling [107].

2. As it was shown in the experimental works of Yu.N. Tyurin and A.P. Vasilevich [93] and also S.A. Petrovsky and the author of this material [69] a group estimation of a specially (on the basis of a certain algorithm and proximity measure) ear-marked sub-group of appraisals is the most often nearer to the truth than a group estimation of the entire group. This trend looks like to be extremely perspective.

3. Within a theoretical aspect the expert information processing is most developed in the sphere of aggregating preferences. Hundreds of foreign works are devoted to this sphere, and out of them let's mention the works of P.S. Laplas, F. Galton, Ch. Dodgeson, E. Cermelo, K.J. Arrow, J. Kemeny, M. Kendall, F. Mosteller, R. Hewbert, and out of the Russian works - the research work of P.Yu. Ghebotarev [103]. The above-mentioned authors consider the problems of aggregating preferences on the basis of a heuristic, axiomatic, optimizational, statistical and a number of other approaches.

 

4. On the opinion of B.G Litvak [49], "one of the basic instruments used in analyzing and processing an expert information are the proximity measures. They allow to define "how close or far are the expert’s points of view." For the first time, the axiomatic proximity measure was introduced in work of J. Kemeny [158]. The proximity measures are widely used in processing the expert information at least in two important aspects:

¾ First of all, they are used for ear-marking "small" and the so called "nuclear" sub-groups from a group of the expert appraisals;

¾ Secondly, the proximity measures are used for aggregating of the expert information, and this is one of the most important problems of the expert appraisal processing.

5. The availability of the computer program support is the most important help for the analysis and processing of the expert information. As N.N. Moiseyev [60] noted "...an intensive progress of both the system analysis and expert prognostication fell on the epoch of an impetuous progress of computers. And later on they will possibly exert a sufficient influence upon the quality of the expert prognostication".