Abstract Four varieties of red garlic (Allium sativum L.) cultivated in different Italian territories, Sulmona (Abruzzo), Proceno and Castelliri (Lazio), and Nubia (Sicily), were analysed by Attenuated Total Reflectance-Fourier Transformed Infrared (ATR-FTIR) spectroscopy. ATR-FTIR spectra of bulbils and bulbil tunics were separately acquired and processed by Partial Least Squares Discriminant Analysis (PLS-DA) with the aim of classifying the garlic samples on the basis of their geographical origin. Finally, two multi-block strategies (based on Sequential and Orthogonalized Partial Least Squares and Sequential and Orthogonalized Covariance Selection, coupled with Fisher’s Linear Discriminant Analysis) have been applied in order to test whether a joint analysis of data could lead to higher prediction rates. Eventually, the best results were achieved by the multi-block approach based on SO-PLS, which allows obtaining a total classification rate of 95 % (corresponding to one misclassified sample over 20) in external validation.

中文翻译：

---

使用快速和非侵入性衰减全反射傅里叶变换红外 (ATR-FTIR) 光谱结合化学计量学对红大蒜 (Allium sativum L.) 进行地理区分

摘要 采用衰减全反射-傅里叶变换红外光谱 (ATR-FTIR) 分析了意大利不同地区苏尔莫纳 (阿布鲁佐)、普罗塞诺和卡斯特利里 (拉齐奥) 和努比亚 (西西里) 种植的四种红蒜 (Allium sativum L.) ) 光谱学。通过偏最小二乘判别分析 (PLS-DA) 分别获取和处理球茎和球茎外衣的 ATR-FTIR 光谱，目的是根据大蒜样品的地理来源对其进行分类。最后，应用了两种多块策略（基于顺序和正交化偏最小二乘法和顺序和正交化协方差选择，加上 Fisher 线性判别分析）来测试数据的联合分析是否可以导致更高的预测率。最终，