Statistics Glossary → Biplot
Biplotconcept
A chart that shows observations (or row categories) and variables (or column categories) together in the same two-dimensional plane.
In a principal component biplot, observations are drawn as points using their component scores, and variables as arrows from the origin using their loadings. The direction of an arrow shows which component a variable is associated with, and the angle between two arrows approximates the correlation between the variables (a narrow angle means strong positive correlation, a right angle no correlation, opposite directions a negative one). The projection of an observation onto an arrow's direction approximates that observation's relative value on that variable. In correspondence analysis, row and column categories are shown on the same map, and categories close to each other indicate similar profiles.
A biplot is a projection of multidimensional data onto just two dimensions; if the first two components explain only a small share of the total variance, distances and angles can mislead. Whether observations or variables are represented faithfully depends on the scaling chosen (the form biplot preserves distances between observations better, the covariance biplot relationships between variables). In correspondence analysis, the distance between a row point and a column point cannot be read directly under every scaling; interpretation usually relies on directions and on distances between points of the same type.
An everyday example
A food magazine examines 20 breakfast cereals on sugar, fibre, protein and calories. In the biplot the sugar and calorie arrows point the same way, while the fibre arrow points the opposite way; chocolate cereals cluster at the tip of the sugar arrow and bran cereals at the tip of the fibre arrow. In a single chart the reader sees a summary of both the products and the relationships between the nutritional values.
In MerQur
In principal component analysis, observation scores and variable loadings are plotted together; in correspondence analysis (CA / MCA), row and column categories are shown on the same map.
Analyses where this concept appears
Correspondence Analysis (CA / MCA)Principal Component Analysis (PCA)