![]() |
Missings |
|
| Are | ||
| Now | ||
| Equally | ||
| Treated |
A brief summary of MANET features
Option-click in any part of any plot provides information specific to that display: counts, values, numbers highlighted. To indicate, that a subsequent click will pop-up an information box, the shape of the mouse changes to a pointer with an additional questionmark, while pressing the option-key.
Examples:
| Querying in a Scatterplot | Querying in a Mosaic Plot |
![]() |
![]() |
The following table lists all possibilities of queries available
in MANET:
(Clicking always means option-clicking)
| Missing Value Chart | Clicking a horizontal bar of the chart shows the number of recorded and missing values in the bar and the numbers highlighted |
| Barchart | Clicking a bar of a barchart shows the number of values in the bar and the numbers highlighted |
| Histogram | Clicking a bar in the histogram shows the number of values in the bar and the numbers highlighted and the limits of the interval of that bar |
| Boxplot | Clicking any part of the box shows the values of the median, upper and lower hinges as well as inner and outer fences |
| Clicking outside the box shows a horizontal rule, labeled with current position of the mouse, and the number of points at this position | |
| Scatterplot | Clicking an axis shows a horizontal respectively a vertical line together with its position in the scatterplot |
| Clicking any horizontal bar at the bottom of the scatterplot shows the number of recorded and missing values in this bar and the numbers highlighted | |
| Clicking on any point in the scatterplot, shows the values (or ranges) of the two coordinates, and the number point at that position (if points overlap) | |
| Mosaic Plot | Clicking any bin in the plot shows a list of the variables and their values corresponding to that bin as well as the number of values falling in this bin and the numbers highlighted |
| Weighted Histogram, Barchart & Mosaic Plot | The interrogation results include the percentage of the weightings in the class and the percentage highlighted |
For scatterplots and polygon plots exists an extended querying. By holding the shift button during the querying all currently selected variables in the variables list are listed with the corresponding values.
MANET has a number of features which become apparent whenever the cursor changes shape as a cue that clicking or clicking and moving will alter the display.
|
Plot |
Cue |
Functionality |
|---|---|---|
| Histogram | ![]() |
Change the bin width of the histogram |
![]() |
Change the offset of the histogram | |
|
Cyclical switching between:
|
||
![]() |
Adjust scaling after a redefinition of the bin width, to fit into the window again. | |
![]() |
Switch to the corresponding spinogram (width proportional to cases for each bin).
|
|
| Barchart | ![]() |
Switch to the corresponding spineplot (width
proportional to cases in that category). Changes to |
![]() |
Grab (while pressing the control key) a bar to exchange or to move to an other position | |
| Dotplot | ![]() |
Cues to switch between a horizontal and a vertical layout of the dotplot. |
| Scatterplot | ![]() |
Flip the x and y axes of the scatterplot |
| Mosaic Plot |
Switch to model display:
|
Indicates hidden information due to
Limited screen resolution
A warning can be found in the context of highlighting. There are two special cases, when highlighting can give trouble. The first case occurs, whenever a very small bar or bin is highlighted with very few values. If the unhighlighted bar or bin is only 2 to 5 pixels of height on the screen, a proportion of less than 10% would have less than 1 pixel of height, and thus not be plotted at all. Although the user could interrogate the bar or bin by a single optionclick, finding out the right values, it is important to prevent the user from overlooking the highlighted values. This is done by a redunderlining of the bar or bin. The complementary case, where nearly all values are highlighted except for a few, not plotted, values, is handled in the same way.
The third situation where warnings can occur, are plots where the binsize differs dramatically from the number of values that bin should represent. This is due to limited resolution of the computer screen.
|
Left red line: Although 100% of the pixels of the bar are highlighted, not all of the cases are selected |
![]() |
Right red line: The highlighting display is correct, but the area of the bar differs dramatically from the size it should have. |
Points falling out of the plot canvas
An often overlooked problem arises if observations fall out of the plot area, and thus are clipped. To indicate this MANET plots an extra red frame inside the plot window.
Application of interactive graphics for advanced selection techniques and graphical database queries.
Packages offering interactive statistical graphics like Data Desk, JMP or SAS-Insight, have been around for about ten years now. These packages support a selection of a subset of the data by different tools and selection modes (and, or, xor, etc.). Usually a selection is directed and hierarchical, what the above mentioned packages do not take into account. Thus changing any element inside a sequence results in a complete redefinition of all selections. Selection-Sequences store all selections in parametric form in a hierarchical object, enabling local changes of any plot in the sequence. In addition the parametric information of Selection Sequences offer an easy way to translate a selection into a database query language like SQL, as well as the handling of relational data.
An Example:
|
Imagine the following scenario: You are interested in the 2-dimensional distribution of LHead vs. LChest - the lower right scatterplot - for different years and protection systems. A natural way to do this, would be to select a year and successively intersect with a protection system. But switching to other years or protection systems would require a totally new selection, starting with the year again. To avoid this, Selection-Sequences store the whole selection path, i.e.
for each window, and add a small icon at the lower right of
each window in the sequence. Thus the user can easily select
another year, or another protection system, and watch the changing
of the scatterplot, without loosing the selection information
in the other plots. |
![]() |
![]() |