Ying Pang: actor-augmented modeling and forecasting: district ungulate abundance

We analyze ungulate abundance of different districts within Kruger National Park, South Africa. The utilization of a disaggregated data set strengthens and extends the understanding about population density and dynamics in regions. We employ a factor model to investigate common movement and heterogeneity for the regional population to specified species. And, we believe that a handful of factors can successfully represent covariate and idiosyncratic information of the data set, which can be consistently estimated using principal components.
We also consider modeling with these multi-level factors as augmented predictors to produce one-step-ahead out of sample forecast estimations. We evaluate the forecasting accuracy and conclude that involving disaggregated factors, which are species- and/or region-specific wide, can be beneficial to improve the predictive performance, however, the improvement to which extent differs across species and regions. Both simulation and empirical studies are demonstrated to provide the evidence.