An inhomogeneous continuous-time Markov chain model is proposed to quantify animal preference and avoidance behaviour in a choice experiment. We develop and apply our model to a choice flume experiment designed to assess the preference or avoidance responses of sea bass (Dicentrarchus labrax) exposed to chlorinated seawater. Due to observed fluctuations in chlorine levels, a stochastic process was applied to describe and account for uncertainty in chlorine concentrations. A hierarchical model was implemented to account for differences between eight experimental runs and use Bayesian methods to quantify preference/avoidance after accounting for observed shoaling behaviour. The application of our method not only overcomes the need to track individuals during an experiment but also circumvents temporal autocorrelation and any violations of independence. Our model therefore surpasses current methods in choice chamber studies, incorporating variability in the environment and group-level dynamics to yield results that scale and generalise to the real-world.

中文翻译：

---

连续时间马尔可夫链模型在选择水槽实验分析中的应用

提出了一种非齐次连续时间马尔可夫链模型来量化选择实验中的动物偏好和回避行为。我们开发并将我们的模型应用于一个选择水槽实验，该实验旨在评估鲈鱼的偏好或回避反应（Dicentrarchus labrax) 暴露于氯化海水中。由于观察到氯含量的波动，应用随机过程来描述和解释氯浓度的不确定性。实施分层模型以解释八次实验运行之间的差异，并在解释观察到的浅滩行为后使用贝叶斯方法量化偏好/回避。我们方法的应用不仅克服了在实验过程中跟踪个体的需要，而且还避免了时间自相关和任何违反独立性的情况。因此，我们的模型在选择室研究中超越了当前的方法，结合了环境和群体级动态的可变性，以产生可扩展和推广到现实世界的结果。