I have read some papers about metal accumulation and metal related biomarkers in marine bivalves from different locations. Authors tend to apply PCA or Discriminant Canonical Analysis to analyze the metal accumulation and biomarker datasets to explain that bivalves from different locations have different metal accumulation and biomarker pattern.
For me, using PCA or DCA analyis results in similar outcome. What is the difference between PCA AND DCA, which one should I use?
Thanks for your answer
Hello Dr. Zhou,
Principal component analysis or PCA is employed to transform a set of variables of apparent correlation into a set of uncorrelated variables in a linear fashion so that the principal components or the set of linear correlated variables are arranged in a hierarchy in terms of having the greatest possible variance and automatically dominate the data sets.
Discriminant canonical analysis is in actuality Linear Discriminant Analysis and is related to both the Canonical correlation and the PCA. It derives interval variables from a nominal variable in sets of data and establishes the canonical correlation with the intervals in a manner which has the variable with highest multiple correlation first, followed by lesser correlated interval variables.
In essence both the statistical analyses point towards more or less the same fact and by scientists other than statisticians they are used interchangeably in general.
The following link should help you in understanding them better:
Regards,
Dr. Abhishek Mukherjee
Hello, if you want to analyse your data by DCA you have to define groups or classes a priori, then the variation caused by the variables will be explained for each group or class that you had detected before the analysis.
For PCA you don't know how the groups or classes are integrated, you have a set of variables and suspect there will be grouping or classification after the analysis according to the influence of variables.
Both PCA and DCA fall into the unconstrained ordination techniques that are used to describe environmental or ecological gradients of variation among individual sampling entities. The key difference is that PCA assumes linear relationship to the gradients while DCA assumes unimodal relationship.
I think you may find this publication very helpful and it should provide you with the answer:
http://cc.oulu.fi/~jarioksa/opetus/metodi/notes.pdf
Cheers,
Tomas
Hi,
I am attaching another paper that you should find very useful.
Tomas
Dear Mariana Díaz-Santana-Iturrios, you mean that in DCA, we have defined groups before the analysis, while in PCA we don't know about the integration of groups prior. But after analysis, we suspect grouping based on the analyzed variables. Interesting and easily understandable.
Thank you
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts