My data comprises of bulk geochemical i.e.,major, trace and REE parameters of granitoids. Mineralogical and norm calculation do not show much variation of data. However, I can correlate data generated on sheared vs. non sheared portion of the granitoids. Please give me hint how to manage data or simply should I manage only the REE data with analysis of variance between sheared/non sheared results. I am undertaking the task of profiling of rare earth elements in the pluton.
Hello Dilip.
Before any geochemical study, it is necessary to interact with "closure problem". You may use "log-ratio transforms" to open the data.
...
Mahalanobis distances method is a non-structural method to identify anomalous samples in multi-element geochemical data.
It provide the degree of anomaly for samples "considering all analysed elements" not a single element.
The main weakness of Mahalanobis distances method is that the model doesn't identify importance of elements in anomalous samples. In other word, you won't find out which element(s) are making the sample anomalous.
as Hossein mentioned, you have compositional data und you might consider the special nature of compositional data, e.g. using the mentioned log-ratio transformations. Mahalanobis distances are itself influenced by outliers, but only if you use a classical estimate of the MD. There are robust alternatives available. You may have a look on R package robCompositions.
- Badges
- Science topic