A dynamic probabilistic principal components model for the analysis of longitudinal metabolomics data
G.Nyamundanda, I.C.Gormley and L.Brennan
Appl. Statist., 63 (2014), 763 -- 782


Programs:
----------
It is unfortunately not possible to distribute the data discussed in the paper.
However, three programs are provided which can be used to replicate the methodology employed in Section 5.1, Section 5.2 and Section 5.4 of the paper.
The programs consist of R code, and are commented throughout as a guide to the practitioner. Full details of the required format of the data are provided within each program.


Section5.1_DPPCA&MetabolomicTrajectories.txt
-------------------------------------------
This program contains R code to perform the analysis of Section 5.1 i.e. the code performs DPPCA and plots the resulting metabolomic trajectories in the principal subspace from the first time point.

Section5.2_DPPCA&LMMs.txt
------------------------
This program contains R code to perform the analysis of Section 5.2 i.e. the code (i) performs DPPCA (ii) highlights influential metabolites and (iii)employs linear mixed models to identify those influential metabolites which change over time. This program calls the script LMM_GibbsAlgorithm.r and so the script LMM_GibbsAlgorithm.r must be saved in the working directory in R. 

Section5.4_AssessingModelFit.txt
------------------------------
This program contains R code to perform the analysis of Section 5.4 i.e. the code implements the posterior predictive model checking approach detailed in the paper through the computation of mean absolute deviations between the observed covariance matrix and that of replicated data sets.


Claire Gormley
School of Mathematical Sciences
University College Dublin
Dublin, Republic of Ireland

E-mail: claire.gormley@ucd.ie

