摘要
为进一步提高烟气水分标准曲线的回归系数和测量准确度,以水分为待测物,以酒精为内标物,以异丙醇为溶剂,进行气相色谱分析。在线性范围内增大标准溶液中水分浓度取值范围即增大自变量的取值范围,对其一元线性回归模型进行优化,以增大回归系数和回归系数的t检验值,分析待测样品的准确度提高。对回归模型的优化进行数理解析,先进行一个常量的证明,然后用差减法和求导法证明自变量的取值范围增大后回归系数和回归系数的t检验值增大,最后推导了回归系数的预测值,并对上面的证明和推导进行了数学验证。回归模型优化要求仪器的响应值即因变量要符合正态分布,在线性范围内仪器响应值的方差是与样品含量或浓度无关的常数,要能精准地控制标准样品含量或标准溶液浓度,标准样品或标准溶液的个数要尽可能的多,如不能满足上述条件只能进行粗略的估计预测。
To further enhance the regression coefficients and measurement accuracy of gas water standard curve,gas chro- matography is carried with water as the analytes,with alcohol as the internal standard,with isopropyl alcohol as solvent. By linear increasing within the concentration range of water in standard solution, namely increasing from the range of inde- pendent variables, one linear regression model has been opti- mized and the regression coefficient and regression coefficient t-test has been increased. Detection and analysis the accuracy of the sample under test is raised. Regression model for opti- mization is mathematical analyzed step by step. First it is to prove a constant; secondly, the regression coefficients and t-test value increases after the range of independent variables increased by subtraction and derivation;finally, prediction of regression coefficients is derived, and the above derivation of a mathematical proof and validation. Regression model optimi- zation requires the value of the instrument response namelydependent variable that should be consistent with normal dis- tribution, the linear range of the value of the variance within the instrument response that should be nothing to do with the content of the sample or the concentration of the constants, to be able to precisely control the standard content of the sampleor standard solution concentration, the standard number of samples or standard solution that should be as much as possi- ble. It can only forecast a rough estimate on the above conditions that are not met.
出处
《化学试剂》
CAS
CSCD
北大核心
2013年第10期915-920,950,共7页
Chemical Reagents
关键词
自变量
因变量
回归系数
气相色谱
化学计量学
independent variable
dependent variable
regres- sion coefficient
GC
chemometrics
