Dan Jensen

A multivariate dynamic linear model for early warnings of diarrhea and pen fouling in slaughter pigs

Animal and environment data are described with a multivariate dynamic linear.Observations repeatedly deviating from model forecasts indicate undesired events.Early and accurate warnings of diarrhea and pen fouling were achieved.The effect of each variable is assessed by omission and inclusion in the model. We present a method for providing early, but indiscriminant, predictions of diarrhea and pen fouling in grower/finisher pigs. We collected data on dispensed feed amount, water flow, drinking bouts frequency, temperature at two positions per pen, and section level humidity from 12 pens (6 double pens) over three full growth periods. The separate data series were co-modeled at pen level with time steps of one hour, using a multivariate dynamic linear model. The step-wise forecast errors of the model were unified using Cholesky decomposition. An alarm was raised if the unified error exceeded a set threshold a sufficient number of times, consecutively. Using this method with a 7day prediction window, we achieved an area under the receiver operating characteristics curve of 0.84. Shorter prediction windows yielded lower performances, but longer prediction windows did not affect the performance.

Describing temporal variation in reticuloruminal pH using continuous monitoring data

Reticuloruminal pH has been linked to subclinical disease in dairy cattle, leading to considerable interest in identifying pH observations below a given threshold. The relatively recent availability of continuously monitored data from pH boluses gives new opportunities for characterizing the normal patterns of pH over time and distinguishing these from abnormal patterns using more sensitive and specific methods than simple thresholds. We fitted a series of statistical models to continuously monitored data from 93 animals on 13 farms to characterize normal variation within and between animals. We used a subset of the data to relate deviations from the normal pattern to the productivity of 24 dairy cows from a single herd. Our findings show substantial variation in pH characteristics between animals, although animals within the same farm tended to show more consistent patterns. There was strong evidence for a predictable diurnal variation in all animals, and up to 70% of the observed variation in pH could be explained using a simple statistical model. For the 24 animals with available production information, there was also a strong association between productivity (as measured by both milk yield and dry matter intake) and deviations from the expected diurnal pattern of pH 2 d before the productivity observation. In contrast, there was no association between productivity and the occurrence of observations below a threshold pH. We conclude that statistical models can be used to account for a substantial proportion of the observed variability in pH and that future work with continuously monitored pH data should focus on deviations from a predictable pattern rather than the frequency of observations below an arbitrary pH threshold.

A simulation study to evaluate the performance of five statistical monitoring methods when applied to different time-series components in the context of control programs for endemic diseases

Disease monitoring and surveillance play a crucial role in control and eradication programs, as it is important to track implemented strategies in order to reduce and/or eliminate a specific disease. The objectives of this study were to assess the performance of different statistical monitoring methods for endemic disease control program scenarios, and to explore what impact of variation (noise) in the data had on the performance of these monitoring methods. We simulated 16 different scenarios of changes in weekly sero-prevalence. The changes included different combinations of increases, decreases and constant sero-prevalence levels (referred as events). Two space-state models were used to model the time series, and different statistical monitoring methods (such as univariate process control algorithms–Shewart Control Chart, Tabular Cumulative Sums, and the V-mask- and monitoring of the trend component–based on 99% confidence intervals and the trend sign) were tested. Performance was evaluated based on the number of iterations in which an alarm was raised for a given week after the changes were introduced. Results revealed that the Shewhart Control Chart was better at detecting increases over decreases in sero-prevalence, whereas the opposite was observed for the Tabular Cumulative Sums. The trend-based methods detected the first event well, but performance was poorer when adapting to several consecutive events. The V-Mask method seemed to perform most consistently, and the impact of noise in the baseline was greater for the Shewhart Control Chart and Tabular Cumulative Sums than for the V-Mask and trend-based methods. The performance of the different statistical monitoring methods varied when monitoring increases and decreases in disease sero-prevalence. Combining two of more methods might improve the potential scope of surveillance systems, allowing them to fulfill different objectives due to their complementary advantages.