On your data

Mapping political space

What you put in:

Any political attitudes you want, measured on the same set of people. Specifically, you provide a table with a column for each question asked of a group of people, and a row for each person's answers. If these questions are political and the answers are numbers, you'll get a map of political space. Of course, you could do this with any kind of data.

How it builds the map:

The resulting map is not built from any preconceived ideas about the shape of political space. Instead, it's simply based on correlations between the variables you provide. If people who tend to score above average on one also score above average on another (in other words, they're positively correlated), those two variables will move toward each other in the map. And if two people who score above average on one variable tend to score below average on another (indicating a negative correlation), those two variables will move apart from each other in the map. Technical details here.

What it communicates

In general, variables that are near each other go together. However, it's important to dig deeper and ask why each variable is where it is. Any visual map is necessarily an imperfect representation of the true underlying network of correlations between variables. This is why CorrExplore's interactive and animated features are so important. These features are designed to communicate the correlation network instead of oversimplifying it like cluster analysis, factor analysis, or other mapping techniques that produce still images.

Interact with the map to discover:

Why variables are positioned where they are (by hovering your mouse over them to see their correlations; non-interactive maps can't do this because they'd have to show all correlations, making the display too cluttered).
Whether a variable has an alternative resting place (by dragging and dropping).
What a cluster like the political left looks like on its own, without reference to the political right (by hiding variables).
How different sets of people see political space differently, for example those with high interest in politics vs low (by using segmentation).
How political space is changing over time or in response to events (by using segmentation on a time variable).
An overview that suggests ideas for other multivariate analyses you might not have thought of.

Built-in political demo data

CorrExplore has a built-in demo data set of political attitudes from the 2004 American National Election Study. Click here for an example analysis based on this data set, or download CorrExplore and try it out yourself.

Visually explore correlation networks