DataLab is a compact statistics package aimed at exploratory data analysis. Please visit the DataLab Web site for more information....



Guided Tour: 3D Rotation

In the following section, we'll have a look at two other methods to inspect data. First, let's have a closer look at the 3D-rotation of data. When it comes to analyzing multidimensional data, human beings exhibit severe restrictions as they cannot imagine more than about four dimensions (e.g. three dimensions in Euclidian space, and the temperature of the data points). This narrow window to the world of multivariate data is further restricted by the two-dimensional nature of computer displays; at the best a three-dimensional plot can be produced on screen by using, for example, color coding to display the third dimension. Thus it is desirable to get another dimension on the computer display. And one way to achieve this is to use on-line rotation of the data using a three-dimensional sub-space projected onto the two dimensions of the computer display. The rotation creates the illusion of a three-dimensional image, which results in a four-dimensional representation if color coding is used for another dimension.

Before looking at 3D-rotation a fair warning should be issued at this point: 3D-rotation requires a lot of computing power and produces it's desired effect only on a suitable hardware (high-frequency CPU, fast graphics card).

In order to start up the 3D-rotation, load the file IRIS.IDT from the disk and then click the command Visualization/3D-Rotation. At first it shows a static display of three variables projected onto the screen. Two modes of rotating displays are available: a manual mode (which is the best way to go after you have gained some experience), and an automatic mode.

For a first look click the Auto Rotation button. This starts the automatic 3D-rotation and rotates the selected sub-space according to some scheme which ensures that eventually you will have a look at the data from all possible directions within a certain amount of time. The speed of the rotation can be adjusted by changing the numbers right to the Auto Rotation button.

The 3D-rotation can be set up in various ways. You may change the size of the display, the type of the axes, and the variables used for the display. Some additional options are available with the context menu of the plot windows (right click the 3D plot window).

Next, let's try the manual rotation. The image can be rotated by pressing and holding down the left mouse button and moving the mouse cursor simultaneously. Moving the cursor horizontally rotates the image around the z- axis, moving vertically rotates it around the x-axis. After some exercise to should be able to quickly get a good overlook of the data.

Sometimes, you immediately see some peculiarities of the data (e.g. outliers) by using a rotating image. So, you may want to identify some objects of the data set. Therefore you can use the command Info on the context menu. When active, clicking the 3D plot window will display the data point next to the current mouse position as marked. Information on the selected data point is displayed in the command panel of the 3D window.


Last Update: 2012-Jul-25