Institute of Forensic Medicine, Goethe-University Frankfurt, Kennedyallee 104, 60596, Frankfurt am Main, Germany.
Int J Legal Med. 2013 Jan;127(1):213-23. doi: 10.1007/s00414-012-0675-9. Epub 2012 Feb 28.
Developmental data of juvenile blow flies (Diptera: Calliphoridae) are typically used to calculate the age of immature stages found on or around a corpse and thus to estimate a minimum post-mortem interval (PMI(min)). However, many of those data sets don't take into account that immature blow flies grow in a non-linear fashion. Linear models do not supply a sufficient reliability on age estimates and may even lead to an erroneous determination of the PMI(min). According to the Daubert standard and the need for improvements in forensic science, new statistic tools like smoothing methods and mixed models allow the modelling of non-linear relationships and expand the field of statistical analyses. The present study introduces into the background and application of these statistical techniques by analysing a model which describes the development of the forensically important blow fly Calliphora vicina at different temperatures. The comparison of three statistical methods (linear regression, generalised additive modelling and generalised additive mixed modelling) clearly demonstrates that only the latter provided regression parameters that reflect the data adequately. We focus explicitly on both the exploration of the data--to assure their quality and to show the importance of checking it carefully prior to conducting the statistical tests--and the validation of the resulting models. Hence, we present a common method for evaluating and testing forensic entomological data sets by using for the first time generalised additive mixed models.
幼龄麻蝇（双翅目：Calliphoridae）的发育数据通常用于计算尸体上或周围发现的未成熟阶段的年龄，从而估算最小死后间隔时间（PMI(min)）。然而，许多这些数据集没有考虑到未成熟的麻蝇是以非线性方式生长的。线性模型不能为年龄估计提供足够的可靠性，甚至可能导致 PMI(min)的错误确定。根据 Daubert 标准和法医科学改进的需要，新的统计工具，如平滑方法和混合模型，可以对非线性关系进行建模，并扩展统计分析领域。本研究通过分析描述法医上重要的麻蝇 Calliphora vicina 在不同温度下发育的模型，介绍了这些统计技术的背景和应用。三种统计方法（线性回归、广义加性模型和广义加性混合模型）的比较清楚地表明，只有后者提供了充分反映数据的回归参数。我们明确关注数据的探索——确保其质量，并在进行统计测试之前仔细检查其重要性——以及生成模型的验证。因此，我们提出了一种通过首次使用广义加性混合模型来评估和测试法医昆虫数据集的通用方法。
