 |
| Titel |
Correlation between variables in compositional data |
| VerfasserIn |
Peter Filzmoser, Petra Kynclova, Karel Hron, Clemens Reimann |
| Konferenz |
EGU General Assembly 2017
|
| Medientyp |
Artikel
|
| Sprache |
en
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 19 (2017) |
| Datensatznummer |
250140904
|
| Publikation (Nr.) |
 EGU/EGU2017-4356.pdf |
|
|
|
|
|
| Zusammenfassung |
| In compositional data analysis, the interest is in analyzing relative information rather than
the measured concentration values directly. This can be done by considering the
log-ratios between all pairs of variables, and by constructing an orthonormal basis
describing this information. In the context of correlation analysis, Kynčlová and
others (2017) introduced the concept of symmetric balances, which are specific
coordinates that capture all relative information of two compositional parts of interest to
the remaining parts. The resulting balances are orthonormal, and thus they can
be used for analyses relying on the Euclidean geometry, starting from graphical
inspection to correlation analysis, testing for uncorrelatedness, and robust estimation of
association.
Based on real data, we will demonstrate that already graphical inspection of the balances
provides interesting insights into the data: trends, groups, and associations can be revealed,
and it can be instructive to link such findings to absolute information. In the context of
geochemistry it turns out that correlations based on symmetric balances serve as a sensitive
tool to reveal underlying geochemical processes.
References
Kynčlová, P., Hron, K., Filzmoser, P. (2017). Correlation between compositional parts
based on symmetric balances. Mathematical Geosciences. To appear. |
|
|
|
|
|
|