Guide Guide Manual


			Multi_R_Designer Quick Guide Manual

GRAPHICAL REPRESENTATION

The Plot creator dialog window presents two panels:

Standard panel, where two types of plot can be created:

1) Parity plot:

on the horizontal axis

on the vertical axis

2) XY plot:

Advanced panel, where a quantitative variable is assigned as the discriminate.

Once the variables are selected, a graphic is displayed. If the graphic is a parity plot, the window presents three panels: Parity plot, Error CDF and Statistics.
aa

XY PLOT / PARITY PLOT: Many options are available relatively to graph edition:

Photo camera icon : saves the graphic into an image .jpg file at any given time.
Zoom To Fit : adjusts the axes scale to its original presentation.
Zoom : delimits and zoom a section on the diagram (Warning: If the selection on the graph surpasses or even touches one of the axis, the zooming process will not execute).
Zoom In : reduces the axis scaling.
Zoom Out : enlarges the axis scaling.
Information on selected points : enables the user to select a group of points on the diagram. In the Information on selected points dialog window, the coordinates of the selected points are displayed as well as the exact source (reference, figures, tables).
Edit graphic properties : changes the diagram appearance under four panels: Text, Appearance , Axis and Series.

In the Text panel, the graph main title and axis titles can be modified as well as their font and color.

In the Appearance panel, the graphic background color, the position of the legend and the grid lines color can be modified.

In the Axis panel, the min-max range of the axis scaling can be set up (normal or logarithm) automatically or manually.

In the Series table, the shape and color of series data can be modified.

Discriminate series : two kinds of discrimination can be carried out on the graphics:
Show only valid points : removes or resets the invalid points (hollow symbols on graphics).
Edit envelopes : By default, the RMSE envelopes based on the first correlation/model in the legend are drawn on parity plots. The user can choose in the Edit envelopes dialog window between No envelopes, RMSE envelopes or Relative envelopes (AARE). The RMSE envelopes and Relative envelopes are based on the statistics of the chosen correlation/model from the list given below the No envelopes section.

ERROR CDF: Performance of the models can be visualized on cumulative distribution function error diagrams presented in the Error CDF panel. The best model corresponds to the one presenting a maximum data percentage (on the vertical axis) for low absolute relative error (left part of the horizontal axis). Some options (icons) in the Parity plot panel are also available. Model performance based on the discriminated series

can also be depicted on separate diagrams, if a discrimination was previously done in the Parity plot section.

STATISTICS: Performance of the models is statistically evaluated in a panel divided in three tables: Data point statistics , Relative error statistics and Absolute relative error statistics . Theses statistics are based on calculation where the experimental data are represented on the X-axis and the calculated data on the Y-axis.

In the following definitions x_i is the experimental data and y_i the predicted value.
The mean values are defined as:

and

ALL tables:

Discrimination: The model's name and/or the discriminated series (values if quantitative variable or names if qualitative variable) are given in this section.

DATA POINT STATISTICS table:

RMSE: Square root of the mean square error

a

rho: Correlation coefficient