APPLICATION PACKAGE FOR EXPERIMENTAL DATA PROCESSING

Version 3.0

 The package is intended for processing results of various experiments and scientific investigations by applied statistics and computational mathematics methods in metrology, instrumentation manufacturing, ecology, medicine, physics, biology, chemistry, economics, sociology and others.

The package is designed for a user unfamiliar with special sections of mathematics, applied statistics and programming it is used for personal computers IBM PC compatible.

The merits of the package are: statistical efficiency, user friendliness. The package control is high-automated. The result presentation is most suitable for the user (tables, plots, comments).

The contents and aspects of problems to solve are treated as a result of expert guestioning with due regard for the features of the unprecedented problems raised.

Programs contained in the package treat problems which can be confined to seven aspects.

 1. Computation of statistical numerical values includes:

·        computation of main numerical values of the phenomenon under study;

·        computation of confidence intervals of main numerical values;

·        computation and the histogram plotting;

·        computation of tolerance intervals;

·        checking of dispersion tolerance of variance estimation in several observation groups;

·        checking of dispersion tolerance of mean arithmetic group at different intragroup variance;

·        checking of dispersion tolerance of arithmetic mean by Fisher method.

 2. Computation of dynamic characteristics includes:

·        identification of the first and second order equations of transient processes under different initial conditions;

·        computation of pulse, transmission, transient and amplitude-phase-freguency characteristics.

 3. Identification of functional dependences includes the dependence recovery:

                  ·        y = a×x^b                      (geometrical);

·        y = a×e^bx                (exponential);

·        y = a×ln(b×x)          (logarithmic);

·        y = a×x^b×e^cx            (geometric-exponential);

·        y = a×(1 - e^-bx)      (inverse-exponential);

·        y = a + b×x^c             (geometrical with a free term);

·        y = a + b×e^cx         (exponential with a free term);

·        y = (a+b×x)×e^cx      (linear-exponential);

·        y = h + (a+b×x)×e^cx(linear-exponential with a free term);

·        y = a×x^c×(1 - bx)^d  (product of geometrical);

·        y = a×x^c + b×x^d      (sum of geometrical);

·        y = a×e^cx + b×e^dx   (sum of exponential);

·        y = h + a×x^c + b×x^d       (sum of geometrical with a free term);

·        y = h + a×e^cx + b×e^dx     (sum of exponential with a free term);

·        y = e^cx×(a×cos(wx) + b×sin(wx)) (exponential-sine);

·        y = h + e^cx×(a×cos(wx) + b×sin(wx))      (exponential-sine with a free term);

·                    (polynomial);

·              (geometric-polynomial);

·              (exponential-polynomial);

·           (logarithmic-polynomial);

·                 (periodic);

·              (linear multiple regression).

These algorithms have been developed with regard to nonstationary variance within the observation range considering nonlinearity factor.

 4. Identification of probability distribution densities includes:

·        estimation of unknown parameters of probability distribution densities: normal, uniform, triangular, trapezoidal, antimodal I, antimodal II, truncated Raileigh, chi-square, Student, binomial, Poisson;

·        identification of the above probability distribution densities according to chi-square test;

·        identification of the above probability distribution functions according to Kolmogoroff-Smirnoff and omega-square tests except for Poisson and Bernoulli distribution functions;

·        checking of distribution normality at N<50.

 5. Decision making application tasks include:

·        computation of density functions of suspended particle sizes in liquid medium during dynamic sedimentation process at preset theoretical and empirical distributions;

·        identification of M-dimensional objects;

·        identification of two empirical distributions;

·        detection of intensity variation of the Poisson flow;

·        detection of intensity "spike" of the Poisson flow;

·        checking of many simple hypothesis.

 6. Time series processing includes:

·        computation of multidimensional time series trends and central random series;

·        determination of forecasted values of the temporal series;

·        computation of one-dimensional distribution law of momentary values of truncated realization of standardized random series;

·        computation of auto- and intercovariance (correlation) functions and conduction of total correlation analysis; the latter includes the following tasks: correlation factor, correlation factor homogeneity, correlation indicator, partial correlation, multiplicity correlation factor;

·        computation of auto- and interspectral power densities;

·        trend checking of central random series for stationarity;

·        checking of central random series for stationarity according to the second moments;

·        checking of central random series for stationarity using covariance matrix computation errors.

·        generation of pseudo-random numbers arranged according to normal, uniform, triangular, trapezoidal, antimodal I, antimodal II, truncated Raileigh, chi-square, Student, binomial, Poisson distribution laws;

·        generation of the Poisson flow.

·        generation of normally distributed random vectors;

·        generation of multidimensional normal Markoff processes;

 There is a standard access to each package program (detailed description is given in the User Manual) and the user can insert each of them as a subroutine into a package of his own.

The package workability has been tested in polar modes; the obtained results verify algorithms stability and reliability as well as high accuracy of the values computed.

   Language options are provided for communication with the package.

    There are two versions of the considered package for operational systems MS DOS and WUNDOWS.

   Cost of the package with the copyrights and with all necessary documents (including detailed description of the package): a) for WUNDOWS version is $ 71 999; b) for MS DOS version is $ 14 160.

   Cost of the package for usual user with detailed description of the package: a) for WUNDOWS version is $ 3 250; b) for MS DOS version is $ 720.

 Team manager: K.J.Kachiashvili, Dr. of Tech. Sc., Prof., Member of the International     Academy of Computer Sciences and Systems.

         Address: 2, University st., Tbilisi, 380043, Georgia. I. Vekua Institute of Applied Mathematic of Tbilisi State University. 

                            Tel.(99 532) 23-72-47; Fax: (99 532) 23-72-47.