Interpretation of Biplot

A biplot is a visualization tool commonly used in Principal Component Analysis (PCA) or

Canonical Correspondence Analysis (CCA) to simultaneously display the relationships between
samples (observations) and variables (features).
1. Interpretation of Axes
•Principal Component Axes (PC1, PC2): The two axes in a biplot represent the first two principal
components (PC1 and PC2), which provide the optimal low-dimensional representation of the
data.
•Explained Variance: Each principal component may have a percentage label, such as PC1 (56%),
PC2 (24%), indicating that PC1 explains 56% of the variance in the data, while PC2 explains 24%.
2. Interpretation of Points (Samples)
•Each point represents a sample (observation).
•Samples that are closer together indicate similar characteristics in the principal component
space.
•The distribution of samples along a specific direction suggests that these samples are mainly
influenced by the corresponding variables.
3. Interpretation of Variables (Arrows)
•Arrow Direction: The direction of an arrow indicates the influence of that
variable in that direction.
•Arrow Length: Longer arrows indicate that the variable contributes more to the
principal component.
•Angle Between Arrows:
• Small angle (close to 0°): The variables are positively correlated (similar
trend).
• Right angle (90°): The variables are uncorrelated.
• Opposite direction (180°): The variables are negatively correlated.
4. Relationship Between Samples and Variables
•If sample points are distributed along the direction of a variable’s arrow, it
suggests that those samples have higher values for that variable.
•If a sample is perpendicular to a variable's arrow, that variable has little
influence on that sample.
Example
In a biplot for food nutrition data:
•If the arrows for protein and fat point in a similar direction, it
suggests a positive correlation between them.
•If some food samples are distributed along the high-protein
direction, those samples likely have higher protein content.
•If the arrows for sugar and protein point in opposite directions (180°
apart), it suggests a negative correlation between them.
Conclusion
A biplot provides an intuitive way to understand PCA results by
simultaneously visualizing:
•The similarity between samples
•The correlation between variables
•The influence of variables on samples
In a biplot, the variables are represented by
arrows, which are used to indicate how each
variable contributes to the principal components.
The arrow itself is a vector that reflects the
relationship between the variables and the new axes
(principal components) created by PCA or CCA.
Here's a more detailed explanation of how the
variables (arrows) are defined:
Direction of the Arrow:
• The direction of an arrow represents the influence or contribution of the variable along the
principal component axes (usually the first and second principal components, PC1 and PC2).
• An arrow pointing to a particular direction indicates that the variable is highly related to or
correlated with the data points in that direction (i.e., in the direction of the principal
components).
• The angle between arrows is important for interpreting relationships:
• Small angle (close to 0°): The variables are positively correlated (i.e., they vary
together).
• Right angle (90°): The variables are uncorrelated (i.e., they vary independently of each
other).
• Opposite direction (180°): The variables are negatively correlated (i.e., they vary
inversely to each other).
2.Length of the Arrow:
• The length of the arrow indicates how strongly the
variable contributes to the principal component.
Longer arrows suggest that the variable has a
greater impact on the variance in the data along
that component, while shorter arrows suggest the
opposite.
• The length is proportional to the variance explained
by that variable in the principal component space.
3.Magnitude of the Arrow (Coordinate values):
• Each variable's arrow is placed in the principal
component space based on the coordinates derived
from the principal components (PC1 and PC2). The
coordinates represent how much each variable
contributes to the overall variance captured by these
components.
• A variable that has high values for the first principal
component (PC1) will have an arrow pointing
strongly along that axis, indicating its importance in
explaining the variance of the data.
4.Variable Contribution to PCA:
• The arrows in a biplot show the linear relationships
between variables and the principal components. Each
arrow reflects how much the respective variable contributes
to each principal component (PC1, PC2, etc.).
• The coordinates of the variable's arrow on the plot are
determined by the loading values for the variable on each
principal component. These loading values are part of the
PCA output and show how much each original variable
"loads" onto the principal components.
Summary of Variable (Arrow) Characteristics in a Biplot:
• Direction: Reflects the influence of the variable on the
principal component(s).
• Length: Indicates how much the variable contributes to
explaining the data's variance.
• Angle between Arrows: Shows the correlation between
variables:
• Small angle: Positive correlation.
• Right angle: No correlation.
• Large angle (close to 180°): Negative correlation.
In essence, variables in a biplot are arrows
that provide insight into how each original
variable contributes to the principal
components and how those variables relate
to one another.