### 4.3 More than two variables

Obviously, it is only possible to represent the relationship between two variables in twodimensional space. It is possible to group scatter plots into a matrix to represent all possible pairs of bivariate relationships within a set of observations of several variables. Alternatively, there are multivariate analyses that represent the overall variability in a dataset with many variables as a series of new summary variables, and the results can then be displayed in two dimensions.

In the case of three variables, an isometric three-dimensional plot can be used. Data points are often shown attached to the 'floor' of the plot (so-called 'ball on stick' plots) to provide a better spatial reference for the two dimensional representation of threedimensional space. The three-dimensional equivalent of a regression line is a surface.