Hydrothermal time (HTT) models describe the time course of seed germination for a population of seeds under specific temperature and water potential conditions. The parameters of the HTT model are usually estimated using either a linear regression, non-linear least squares estimation or a generalized linear regression model. There are problems with these approaches, including loss of information, and censoring and lack of independence in the germination data. Model estimation may require optimization, and this can have a heavy computational burden. Here, we compare non-linear regression with survival and Bayesian methods, to estimate HTT models for germination of two clover species. All three methods estimated similar HTT model parameters with similar root mean squared errors. However, the Bayesian approach allowed (1) efficient estimation of model parameters without the need for computation-intensive methods and (2) easy comparison of HTT parameters for the two clover species. HTT models that accounted for a species effect were superior to those that did not. Inspection of credibility intervals and estimated posterior distributions for the Bayesian HTT model shows that it is credible that most HTT model parameters were different for the two clover species, and these differences were consistent with known biological differences between species in their germination behaviour.

中文翻译：

---

三种不同统计方法（非线性最小二乘回归、生存分析和贝叶斯推断）在估计种子萌发热液时间模型中的有用性的比较

水热时间 (HTT) 模型描述了种子群体在特定温度和水势条件下的种子萌发时间过程。HTT 模型的参数通常使用线性回归、非线性最小二乘估计或广义线性回归模型来估计。这些方法存在问题，包括信息丢失、发芽数据审查和缺乏独立性。模型估计可能需要优化，这可能会带来沉重的计算负担。在这里，我们将非线性回归与生存和贝叶斯方法进行比较，以估计两种三叶草物种发芽的 HTT 模型。所有三种方法都估计了具有相似均方根误差的相似 HTT 模型参数。然而，贝叶斯方法允许 (1) 有效估计模型参数而不需要计算密集型方法和 (2) 两种三叶草物种的 HTT 参数的轻松比较。解释了物种效应的 HTT 模型优于那些没有解释的模型。检查贝叶斯 HTT 模型的可信区间和估计的后验分布表明，对于两种三叶草物种，大多数 HTT 模型参数不同是可信的，并且这些差异与物种之间在其萌发行为方面的已知生物学差异是一致的。