Introduction
Multidimensional Scaling is a technique for visualizing distances between objects where the distance is known between the pairs of objects. When the input type is variables, the probability that each point has the same class as its nearest neighbor is calculated. This article describes how to create a a two-dimensional scatterplot, where each of the objects is represented as a point:
Requirements
A Displayr document containing the variables that you want to use as inputs to the Multidimensional Scaling analysis.
Method 1
1. From the menus, select Anything > Advanced Analysis > Dimension Reduction > Multidimensional Scaling (MDS).
2. In the object inspector under Inputs > Algorithm select MDS - Metric or MDS - Non-metric.
Note: In practical application of MDS, there tend to be contradictions in the data and it is impossible to show all the distances on the map accurately. Different MDS algorithms have been developed which make different decisions about how to reconcile these contradictions.
- Metric MDS minimizes the difference between distances in input and output spaces, which attempts to show the distances so that they are, on average, correct.
- Non-metric MDS aims to preserve the ranking or relative ordering of distances between input and output spaces.
Often we care more about relative positioning than absolute differences, in which case non-metric is preferred to metric MDS. For the example below, I've selected Metric.
3. Under Inputs > Variables, select the input variables from your data set.
4. Click the Calculate button to generate the MDS output.
5. OPTIONAL: You can perform color-based grouping in the scatterplot by selecting a variable in the Group variable field.
See Also
How to do Principal Component Analysis in Displayr
How to Save Components/Dimensions from a Dimension Reduction Output
How to Create a Dimension Reduction Scatterplot
How to do Multidimensional Scaling from a Distance Matrix
Comments
0 comments
Article is closed for comments.