WORLD METEOROLOGICAL ORGANIZATION

HOMS COMPONENT

Explanation of the criteria for classification and numbering of components    I73.3.01
(DEC 99)
Explanation of the dates on the component's description

RECONSTITUTION OF MISSING DATA BY REGRESSION (RÉMUS)

 

          

1.   Purpose and objectives

     RÉMUS reconstitutes missing values by using data observed at neighbouring sites. This is done by means of various regression models, including advanced techniques such as ridge regression which reduces the effect of collinearity between explanatory variables.

2.   Description

     To simulate the production of energy at a given site, Hydro-Quebec uses monthly mean values calculated on the basis of daily measurements. However, often some data in the monthly series are missing and must be reconstructed to provide a better basis for simulation studies.

     RÉMUS is designed to reconstitute missing data with advanced modelling tools. The following options are available:

     *   multiple regression;
     *   ridge regression (to avoid the problem of collinearity of
         independent variables); 
     *   forward stepwise regression;
     *   multivariate regression (to account for the correlation between
         several dependent variables);
     *   correlation matrix;
     *   common transformations (1/x, 1/x2, log x, vx, x2) of independent
         variables;
     *   Box-Cox transformation of dependent variable;
     *   graph of Y versus X;
     *   graph of concomitance;
     *   graph of residuals versus predicted values;
     *   graph of residuals versus time;
     *   graph of residuals versus any dependent variable;
     *   graph of residuals on normal paper;
     *   context sensitive help (theoretical and technical);
     *   data reconstitution.

3.   Input

     RÉMUS was originally developed for modelling monthly mean flows, but other variables can be used as well. As default, the input data are assumed to have the format used by Environment Quebec, but the user can specify any other format as long as it is an ASCII format.

4.   Output

     RÉMUS prints on screen or printer:

     *   results from different regression models (cf. 2. Description);
     *   general statistics of variables;
     *   graph of Y versus X;
     *   graph of concomitance;
     *   graph of residuals (cf. 2. Description).

5.   Operational requirements and restrictions

     RÉMUS must be used on an IBM (or 100% compatible) personal computer, with an EGA graphics display (or better), and a hard disk with a minimum of 1 megabyte free. It is designed for use with MS-Dos 3.x (or later).

6.   Form of presentation

     RÉMUS is implemented in Pascal using object oriented programming and is contained on one diskette (3.5", 1.44 MB). The user guide (116 pages) and the theoretical guide (60 pages) are in French. 

7.   Operational experience

     Presently, the software Rémus is being used by several companies and organizations in Canada, and especially by the scientists of Hydro-Quebec.

8.   Originator and technical support

     RÉMUS was developed at INRS-Eau (National Institute of Scientific Research - Water), Quebec, Canada in collaboration with Hydro-Quebec. The project was financed by Hydro-Quebec and the Natural Sciences and Engineering Council of Canada.  Theoretical and technical information can 
be obtained from the following persons: Dr B. Bobée (Tel: 418 654 2539) or Mr H. Perron (Tel: 418 654 3751, Fax: 418 654 2600).

9.   Availability

     From the HOMS National Reference Centre for Canada.

10.  Conditions on use

     The software is available at the price of $ 99.00 (CAN).

 

(First entered: 01 MAR 94

Last updated: 15 DEC 1999)
 top