QDA Miner User’s Manual 159
Interpreting Correspondence Analysis Results
Interpretation of correspondence analysis maps can be somewhat tricky and should be done with great
care, especially when examining the relationship between row points and column points. Here are some
basic rules that should help you interpret these maps:
Relationship among codes (row points)
• The more similar the distribution of a code among subgroups is to the total distribution of all
codes within subgroups, the closer it will be to the origin. Codes that are plotted far from this
point of origin have singular distributions.
• If two codes have similar distributions (or profiles) among subgroups of the independent variable
(columns), their points in the correspondence analysis plot will be close together. For example, if
the codes consist of artist names and the studied subgroups represent different age groups then, if
the form of the distribution of two different artists among these age groups is similar, they will
tend to appear near each other. Codes with different profiles will be plotted far from each other.
Note: Two points may appear close to each other on a two- or three-axis solution, but may in fact
be far apart when an additional dimension is taken into account.
Relationship among subgroups (column points)
• The more singular a profile of coding for a subgroup is, compared with the distribution of these
codes for the entire sample, the farther this subgroup will be from the point of origin.
• If two subgroups of individuals have similar profiles of coding, they will be plotted near each
other. Subgroups with different profiles will be plotted far from each other. Again, it is important
to remember that two points may appear close to each other on a two- or three-axis solution, but
may in fact be far apart when an additional dimension is taken into account.
Relationship between codes and subgroups (row and column points):
• Great caution should be taken when interpreting the distances between two sets of points (row and
column points). The fact that the name of a subgroup is near a specific code should not
necessarily be interpreted as an indication that they are closely related.
• While the distance between codes and subgroups has no interpretable meaning, the angle between
this code point and a subgroup point from the origin is meaningful:
• An acute angle indicates that the two characteristics are correlated.
• An obtuse angle, near 180 degrees, indicates that the two characteristics are negatively correlated.
In the example below, REM, U2 and RadioHead could be viewed as related to 18-year-old listeners.
However, despite the fact that both REM and U2 points are closer to the 18-year-old subgroup,