Comparison of 1-step and 2-step methods of fitting microbiological models
Comparison of 1-step and 2-step methods of fitting microbiological models
复制标题
DOI：
10.1016/j.ijfoodmicro.2012.09.017
复制
发表时间：
2012-11-15
影响因子：
5.4
通讯作者：
Jewell, Keith
中科院分区：
文献类型：
--
作者：
Jewell, Keith
在三个不同的数据集上的应用证实了以前的结论，即一步拟合法比传统的两步法给出的系数更精确。结果还表明，与两步拟合法相比，一步拟合法对具有直接可解释的回归诊断和标准误差的数据提供了更好的拟合度(通常是实质上)。在极端的环境条件下，这种改善是最大的，并且表明一步拟合能指示不适当的函数形式，而两步拟合不能。一步拟合在估计主要参数(如滞后、增长率)以及浓度方面更好，并且数据效率更高，允许在较小的数据集上构建更稳健的模型。一步法可直接应用于可对其使用两步法的任何数据集，此外还可应用于两步法失败的某些数据集。两步法适用于模型开发早期阶段的视觉评估，并且可能是为一步拟合生成起始值的一种方便方法，但任何定量评估都应使用一步法。(C)2012爱思唯尔B.V.保留所有权利。
Previous conclusions that a 1-step fitting method gives more precise coefficients than the traditional 2-step method are confirmed by application to three different data sets. It is also shown that, in comparison to 2-step fits, the 1-step method gives better fits to the data (often substantially) with directly interpretable regression diagnostics and standard errors. The improvement is greatest at extremes of environmental conditions and it is shown that 1-step fits can indicate inappropriate functional forms when 2-step fits do not. 1-step fits are better at estimating primary parameters (e.g. lag, growth rate) as well as concentrations, and are much more data efficient, allowing the construction of more robust models on smaller data sets. The 1-step method can be straightforwardly applied to any data set for which the 2-step method can be used and additionally to some data sets where the 2-step method fails. A 2-step approach is appropriate for visual assessment in the early stages of model development, and may be a convenient way to generate starting values for a 1-step fit, but the 1-step approach should be used for any quantitative assessment. (C) 2012 Elsevier B.V. All rights reserved.