DataLab is a compact statistics package aiming at exploratory data analysis. Please visit the DataLab Web site for more information.... 
Home Features of DataLab Displaying Data Contour Plots  
See also: 3D Rotation, Create a kNN Model


Contour Plots
DataLab enables you to create colorencoded surface charts using three variables. Two variables determine the axes of the diagram and the third variable is used for creating the color surface. The surface charts remind of geographic maps with colorencoded contour lines. Missing data are extrapolated from neighboring data, in order to be able to create a complete color chart for any data. The resolution of the chart determines the duration of the calculation process. The calculation process has to be started by clicking the button Calculate.
You select the variables by clicking the corresponding buttons. If you wish the xycoordinates of the defining data being visible, you have to activate the selection box "Display Data Points". The colors of the contour plot can be set arbitrarily by defining colors for the low range, the middle and the high range of the zvalues. In addition to the default setups "Standard Colors" and "Gray Scale", you can also load color scales from the hard disk (*.cpc files in the DataLab work directory). In addition, the color scale can be adjusted to cover either the total range of the data (Min/Max) or the interquartil range. In most cases "Min/Max" is used, the interquartil range is applied for emphasizing the average data range on the z axis. In the selection box "Weighing Mode", you can determine the way the color chart is calculated by interpolating the data. Using the option "Gaussian", the interpolated value is calculated by the weighted sum of the k nearest neighbor, the weights being derinved from Gaussian functions. The weights depend on the width of the Gaussian function, which can be determined by the parameter "Smoothing". If the "Smoothing" parameter is too low (less than 1.0), all k neighbors will be included in the calculation of the interpolated value resulting in a nonsmooth shape of the interpolated surface. If the "Smoothing" parameter is set too high (over 2.5) only the verfy next neighbors will be included in the calculation which also causes the curve being rugged. Therefore this parameter should be set in the range between 1.3 and 1.7. The option "Average" will cause the interpolated value to be the average of the k nearest neighbors; the option "Median" will assign the median of the k nearest neighbors to the interpolated value. The last option is usually used for classification, while the option "Gaussian" is used for estimating smooth response curves. The parameter "Resolution" determines the number of interpolation points per axis for the calculation of the color chart. The higher this value, the better the resolution of the chart. The time needed to calculate the contour plot increases with the square of this value. The parameter "# Neighbors" determines the number of the neighbors which are included in the calculation of the Z value. The higher this value, the more blurred and averaged the color chart will be. If this parameter is set too low, the interpolation wont be smooth. Five nearest neighbors is a good compromise in many cases. When you move the mouse in the color surface chart, the mouse coordinates will be displayed as the caption of the variable selection buttons. For the zaxis the interpolated value will be displayed (and not the zvalue of the nearest neighbor). 