Abstract Two strategies based on multivariate qualitative and quantitative analysis are proposed for determining whether paprika samples are adulterated with Sudan I or II dyes. The samples were characterized by their UV–visible spectrum. Then, the following chemometric techniques were applied: principal component analysis (PCA) for the data exploration stage, partial least squares-discriminant analysis (PLS-DA) for the qualitative strategy and partial least squares (PLS) for the Sudan quantification strategy. Three classes were considered in the qualitative analysis: unadulterated, adulterated with Sudan I and adulterated with Sudan II. For the quantitative analysis, models were developed for each adulterant in a concentration range of 0–12 mg g−1, where 0 mg g−1 stands for the unadulterated samples. Both strategies gave satisfactory results. The multi-class PLS-DA model properly differentiate between the three types of studied samples: non-adulterated and adulterated with Sudan I and with Sudan II. Only samples at the lowest concentration (1 mg g−1) were wrongly detected as unadulterated. The PLS models were validated by setting the regression line of the model-predicted concentration versus the real concentration for the test set samples. The slope and intercept of each regression line were compared with those of the ideal line (0, 1) by means of the joint test of ordinate and slope.

中文翻译：

---

用 SUDAN I 和 II 染料确定辣椒粉掺假的定性和定量多变量策略

摘要 提出了两种基于多变量定性和定量分析的策略来确定辣椒粉样品是否掺有苏丹 I 或 II 染料。样品通过其紫外-可见光谱表征。然后，应用了以下化学计量技术：数据探索阶段的主成分分析 (PCA)、定性策略的偏最小二乘判别分析 (PLS-DA) 和苏丹量化策略的偏最小二乘法 (PLS)。在定性分析中考虑了三类：未掺假、掺有苏丹 I 和掺有苏丹 II。对于定量分析，我们为每种掺杂物在 0–12 mg g-1 的浓度范围内开发了模型，其中 0 mg g-1 代表未掺杂的样品。这两种策略都取得了令人满意的结果。多类 PLS-DA 模型正确区分了三种类型的研究样本：未掺假和掺有苏丹 I 和苏丹 II。只有最低浓度 (1 mg g-1) 的样品被错误地检测为纯正样品。通过设置模型预测浓度与测试集样品的实际浓度的回归线来验证 PLS 模型。通过纵坐标和斜率的联合检验，将每条回归线的斜率和截距与理想线(0, 1)的斜率和截距进行比较。通过设置模型预测浓度与测试集样品的实际浓度的回归线来验证 PLS 模型。通过纵坐标和斜率的联合检验，将每条回归线的斜率和截距与理想线(0, 1)的斜率和截距进行比较。通过设置模型预测浓度与测试集样品的实际浓度的回归线来验证 PLS 模型。通过纵坐标和斜率的联合检验，将每条回归线的斜率和截距与理想线(0, 1)的斜率和截距进行比较。