For the selectivity evaluation of core-shell silica, siloxane-bonded reversed-phase columns by the solvation parameter model a minimal set of thirty-five calibration compounds were identified for applications with mobile phases containing from 20-70 % (v/v) methanol- or acetonitrile-water. The Kennard-Stone uniform mapping algorithm was employed to select the calibration compounds from a larger database of compounds with known retention properties used previously for column selectivity evaluation. The number and identity of the calibration compounds was optimized by considering the statistical quality of the calibration models, the absolute value for the standard deviation of the individual system constants, and the closeness of the estimated system constants for the calibration models to the values for the full data set on seven core-shell columns with different morphology for mobile phases containing 50 % (v/v) methanol- or acetonitrile-water. The predictive capability of the calibration models was evaluated for an external test set for 50 % (v/v) methanol-water for the same columns with a typical root mean square error of prediction (log k) of 0.028 for the thirty-five calibration compounds (n = 35) models. For all columns and mobile phase compositions the calibration models with n = 35 had a range for the coefficient of determination from 0.999 to 0.970 with < 5 % smaller than 0.99 (84 models) and a standard error of the estimate from 0.011 to 0.057 with 75 % < 0.030. System constants determined at 10 % (v/v) increments of organic solvent were utilized for the construction of system maps and correlation diagrams on Kinetex columns with octadecylsiloxane-bonded (C18 and EVO C18), diisobutyloctadecylsiloxane-bonded (XB-C18) octylsiloxane-bonded (C8), biphenylsiloxane-bonded (Biphenyl), phenylhexylsiloxane-bonded (Phenyl-Hexyl) and pentafluorophenylpropylsiloxane-bonded (F5) silica stationary phases for 20-70 % (v/v) methanol- and acetonitrile-water mobile phases. System maps for the Kinetex Phenyl-Hexyl column and correlation diagrams for columns of similar selectivity (Kinetex C8 and Phenyl-Hexyl) and dissimilar selectivity (Kinetex EVO C18 and Biphenyl) is used to illustrate the general approach to selectivity evaluation.

为了通过溶剂化参数模型评估核-壳二氧化硅，硅氧烷键合的反相色谱柱的选择性，鉴定出最少的35种校准化合物，用于含20-70％（v / v）甲醇的流动相-或乙腈-水。Kennard-Stone均匀映射算法用于从较大的具有已知保留特性的化合物数据库中选择校正化合物，该数据库以前用于色谱柱选择性评估。通过考虑校准模型的统计质量，各个系统常数的标准偏差的绝对值，可以优化校准化合物的数量和特性。校准模型的估计系统常数与包含7种核-壳型色谱柱的完整数据集的值的接近度，这些色谱柱具有不同形态的流动相，其中包含50％（v / v）甲醇或乙腈-水。对于相同色谱柱的50％（v / v）甲醇-水的外部测试装置，对校准模型的预测能力进行了评估，对于35个校准，典型的预测均方根误差（log k）为0.028化合物（n  = 35）个模型。对于所有列和流动相组成校正模型与ñ = 35的确定系数范围为0.999至0.970（小于5％小于0.99（84个模型）），估计的标准误为0.011至0.057（75％<0.030）。以有机溶剂的10％（v / v）增量确定的系统常数用于在带有十八烷基硅氧烷键（C18和EVO C18），二异丁基十八烷基硅氧烷键（XB-C18）辛基硅氧烷-的Kinetex色谱柱上构建系统图和相关图键合（C8），联苯硅氧烷键合（Biphenyl），苯己基硅氧烷键合（苯基-己基）和五氟苯基丙基硅氧烷键合（F5）硅胶固定相，用于20-70％（v / v）甲醇和乙腈-水流动相。