Using state-space models to predict the abundance of juvenile and adult sea lice on Atlantic salmon

Estimating the dispersal of Lepeophtheirus salmonis sea lice within and among Atlantic salmon sites of the Bay of Fundy, New Brunswick

The objective of this study was to estimate the impact of infestation pressures on the abundance of the parasitic sea louse, Lepeophtheirus salmonis, in the Bay of Fundy, New Brunswick (NB), Canada, using the Fish-iTrends database for the years 2009-2018. Infestation pressures were calculated as time-lagged weighted averages of the abundance of adult female (AF) sea lice within a site (internal infestation pressure: IIP) and among sites (external infestation pressure: EIP). The EIP weights were calculated from seaway distances among sites and a Gaussian kernel density for bandwidths of 5 to 60 km. The EIP with a bandwidth of 10 km had the best fit, as determined with Akaike's information criterion, and historical AF sea lice abundance. This estimated dispersal distance of 10 km was similar to previous studies in Norway, Scotland and in New Brunswick. The infestation pressures estimated from empirical AF sea lice abundance within and among sites significantly increased the abundance of AF sea lice (p < .001). This study concludes that sea lice burdens within Atlantic salmon farms in the Bay of Fundy, NB, are affected by within site management and could be improved by synchronizing treatments between sites.

A time series approach to player churn and conversion in videogames

Players of a free-to-play game are divided into three main groups: non-paying active users, paying active users and inactive users. A State Space time series approach is then used to model the daily conversion rates between the different groups, i.e., the probability of transitioning from one group to another. This allows, not only for predictions on how these rates are to evolve, but also for a deeper understanding of the impact that in-game planning and calendar effects have. It is also used in this work for the detection of marketing and promotion campaigns about which no information is available. In particular, two different State Space formulations are considered and compared: an Autoregressive Integrated Moving Average process and an Unobserved Components approach, in both cases with a linear regression to explanatory variables. Both yield very close estimations for covariate parameters, producing forecasts with similar performances for most transition rates. While the Unobserved Components approach is more robust and needs less human intervention in regards to model definition, it produces significantly worse forecasts for non-paying user abandonment probability. More critically, it also fails to detect a plausible marketing and promotion campaign scenario.

Estimating sea lice infestation pressure on salmon farms: Comparing different methods using multivariate state-space models

Sea lice are ectoparasites of salmonids, and are considered to be one of the main threats to Atlantic salmon farming. Sea lice infestation on a farm is usually initiated by attachment of the free-living copepodid stage derived from the surrounding water, frequently originating from adult lice on the same farm or from neighboring salmonid farms, referred to as internal and external sources, respectively. Various approaches have been proposed to quantify sea lice infestation pressure on farms to improve the management of this pest. Here, we review and compare five of these methods based on sea lice data from 20 farms located near Grand Manan island in the Bay of Fundy, New Brunswick, Canada. Internal and external infestation pressures (IIP and EIP, respectively) were estimated using different approaches, and their effects were modeled either by a unique parameter for all production cycles or by different parameters for each production cycle, using a multivariate state-space model. Predictive variables, such as water temperature and sea lice treatments, were included in the model, and their effects across production cycles were estimated along with those of other model parameters. Results showed that models with only EIP explained the variation in the data better than models with only IIP, and that models with unique IIP and unique EIP for all cycles were generally associated with the best model fit. The simplest, fixed lag method for calculating infestation pressure had the best predictive performance in our models among the methods studied.