Psychometric Networks for Intensive Longitudinal Data
Description
In the past decade, mental processes have been conceptualized as complex networks of interacting psychiatric symptoms. These networks that can be visualized by means of conditional independence graphs. For estimating the underlying directed graph from intensive longitudinal data, vector autoregression (VAR) is the most commonly used tool. This package wraps several methods in the VAR family that can be used to estimate conditional independence graphs (networks) from multivariate time-series data. The package can fit the simple VAR and VAR with exogenous variables (VARX) model Lutkepohl, H. (2005) that are currently available from the R package 'vars', and its sparse alternative by Basu S. and Michailidis, G.(2015) and sparse vector error correction model (SVECM) implemented in the R package 'sparsevar'. The sparse graphical VAR with covariance estimation by Wild, B., Eichler, M., Friederich, H. C., Hartmann, M., Zipfel, S., & Herzog, W. (2010) from the R package 'graphicalVAR' and the dynamic factor model (DFM) by Doz, Gianone & Reichlin (2011) from the R package 'dynfactoR' are also available. Sparse estimation of high dimensional VAR, VARX, and vector autoregressive moving average (VARMA) and models using hierarchical lag structures Nicholson, W. B., Bien, J., Matteson, D. S. (2017) implemented from the R package 'bigtime' and mixed VAR for symptom time series with marginal distributions in the exponential family Haslbeck, J., Waldorp, L. J. (2015) from the package 'mgm' can also be used with this package. For the inference of symptom networks from multivariate time series of multiple individuals the 'psychNET' package adopts the multi-level VAR (MLVAR) by Epskamp, S., Waldorp, L. J., Mottus, R., & Borsboom, D. (2017) implemented in the R package 'mlVAR' and for the high-dimensional setting the sparse time series chain graphical model by Abegaz, F., Wit, E. (2013) available from the R package 'sparseTSCGM'.