The CLIC package

• Collecting landmarks (button COO)
  Only reference images are downloaded, not the coordinates of their landmarks (see BMC Research Note ). You have to select a set of LM and digitize them. You have to do that on the reference images and, separately, on the images of your own specimens. The relevant output files for subsequent analyses are ending by ..._format.txt; the one ending by ..._DB.txt may be copy-pasted into a spreadsheet to allow the building of a small database.


• Obtaining shape variables (button MOG). The MOG module allows:
  
  • the Generalized Procrustes Analyses, with visual appreciation of each step, producing the residual coordinates (..._ALIGNED.txt) and the Procrustes residuals;
  • the principal component analysis (PCA) of the residual coordinates;
  • the principal component analysis (PCA) of the Procrustes residuals, thus producing the "Procrustes components (PrCp)". For close forms, the "PrCp" can be used as shape variables like partial warps scores (see hereunder);
  • the computation of partial warps scores (PW);
  • the introduction of your own, unknown specimens as supplementary data; use the raw, CLIC formatted coordinates (..._format.txt) (button EXT/UNKN );
  • the computation of the PW on the grand total, i.e. the reference and the unknown specimens;
  • the computation of the PW on each sub-total, i.e. the reference and each unknown specimens;
  • the principal component analysis (PCA) of the PW, producing the relative warps (RW);
  • the Neighbor Joigning tree computed from the Procrustes distances among groups;
  • the allometric content of the two first Procrustes components, RW or canonical variables
  • the discriminant analysis (PCA) on either the Procrustes components or the PW of the reference specimens;
  • and the classification analyzes. The classification procedure is based on the Procrustes distances, on the Mahalanobis distances, as described hereafter...
  
  
• Identifying "unknown" specimens (MOG module)
  
  • Procrustes classification. After the partial warps (PW) have been computed on your input data, a new button appears (EXT/UNK) which allows you to enter external data in the same format (..._format.txt) as the input data. Then a first classification of your external, unknown specimens will be automatically performed on the basis of pairwise Procrustes distances. This sequence is chronological, not logical since Procrustes distances computation does not need the previous transformation of residuals into PW. The Procrustes classification will use two algorithms, one based on the shortest Procrustes distance to each consensus (each unknown with each consensus), and another one based on the K nearest neighbors method (KNN, each unknown with each reference image).
    
  • Mahalanobis classification. After the Procrustes classification of the unknown individual(s), you can ask for a Classification based on discriminant analysis (DA). You will not have the choice of any input or external file. As many times as there are external individuals, the MOG "Mahalanobis classification" successively recreates situations where only one individual has to be classified.
    This is called the "one-by-one" procedure:
    (1) for each individual classification, the PW will be recomputed from the raw reference data and the raw coordinates of the single individual to be classified,
    (2) the Mahalanobis distances will be estimated from these PW, and classified as for a KNN analysis.
    (3) If the reference data have small groups, the set of a few first RW is used as input data instead of the totality of the PW.
  • If you have many unknown individuals, the "one-by-one" procedure can be slow but it is considered as essential for an optimal shape-based classification (see BMC Research Note ).