One document plays two different roles in the corpus document itself and a citation of other documents Bernoulli process topic (BPT) model which considers the corpus at two levels: document level and citation level &latent topic space associated with its roles Multi-level hierarchical structure of citation network is captured by a generative process involving a Bernoulli process Efficient computation algorithm is proposed to overcome the difficulty of matrix inverse operation BPT model to well known corpora to discover the latent topics, recommend important citations, detect the trends of various research areas. The autoregressive (AR) process is a classical model for time series analysis that we will use as a building block. An AR model assumes that each observation is a function of some fixed number of previous observations plus an uncorrelated innovation. Motivated by such goals, in this article we explored a nontraditional treatment of time series analysis by examining models for collections of time series. We proposed a Bayesian nonparametric model for multiple time series based on ideas analogous to Dirichlet-multinomial modeling of documents. We also reviewed a Bayesian nonparametric model based on a beta-Ber noulli framework that directly allows for sparse association of time series with dynamic regimes. Such a model enables decoupling the presence of a dynamic regime from its prevalence.