Compares statistically sequences of states (behavior, texts, molecular data) by likelihood ratio tests on their markovian transition matrices. Performs also a cluster analysis of the sequences and a Principal Coordinates Analysis on the distance matrix between them.
compseq(ser,alpha=0.05,meth="ward.D",printdata=FALSE,printdico=TRUE,printmat=FALSE,
eps=1e-07,clust=TRUE,pca=TRUE)list of list: set of sequences
numeric: global risk threshold for pairwise comparisons.
character:Clustering method. cf hclust.
Boolean:Print original data.
Boolean:Print the dictionnary of states from ser.
Boolean: print all transition matrices and the consensus matrix.
numeric: precision for the convergence of cmdscale.
Boolean: do the cluster analysis.
Boolean: do the principal coordinates analysis.
an object of class compseq with attributes
dico Dictionnary of states
mdist Matrix of pairwise distances between sequences
msign Matrix of pairwise significance levels between sequences
mcom Common or consensus transition matrix
The log likelihood ratio times -2 is used both for tests (Chi-Square approximation followed by Bonferroni post hoc tests) and as a distance to cluster the sequences and to represent them on factorial plans (Principal Coordinates Analysis). Warning: not a metric distance. Susceptible to give incoherent clustering with some methods (meth).
This function does essentially the same work as compmat but with matrices instead of sequences entry.
Pierre, J. S. and C. Kasper (1990). The Design of Ethological Flow-Charts on Factorial Analysis Representations - an Application to the Study of the male Mole-Cricket Sexual Courtship. Biology of Behaviour 15(3-4): 125-151. Van der Heijden, P. G. M. 1986. Transition matrices, model fitting and correspondence analysis. In: Data Analysis and Informatics IV (Ed. by E. Diday), pp. 221-226. Elsevier Science Publishers.
# NOT RUN {
data(seriseq)
compseq(seriseq)
# }
Run the code above in your browser using DataLab