For your first lecture of this guide, I suggest you to skip this section a go directly to section 5. In this section, we will assume that you are familiar with the usage of StarDust.
Inside StarDust, all the algorithms (PCA, K-means & Ward) are able to take into account user-defined “weights” on the columns and on the rows.
A large weight on a row of your dataset means that this particular individual is very important. A large weight on a column of your dataset means that this column is very important. You can set the row-weights here: and the column-weights here: .
The column-weights are designated as in section 5.5.1.
The sum of the row-weights inside a segment is designated as in section 22.214.171.124.
There exist some optimal column-weights that allow you to visualize in an optimal way the characteristics that separate your “Target Group” from the rest of the population. When you are using these optimal column-weights, you can easily “see” and “explore” your “Target Group” in relation to the other individuals.
For example, for the census-income database:
Note that we added as illustrative variable the “taxable income amount”.
A good idea is to also add as illustrative variable the “sex”.
Thereafter, we load the dataset. We go directly to the chart representing the variable “Taxable income amount” (click The button) and we add the “True” modality as a new group.
Let’s configure the display of our Groups: click the button and select the “(Meta-)Group” tab:
... and we finally obtain:
Apparently, in this case, the “Target Group” is roughly divided in two parts! This is interesting... We should investigate what’s the difference between these parts.