Dirksen A, Groth S
Bull Eur Physiopathol Respir. 1986 May-Jun;22(3):231-7.
The powerful computers that are now available enable calculation of reference values by endless permutations of regression models that are not always biologically meaningful. The aim of this study was to test the ability of simple regression models to describe data that has been collected for calculation of reference values. Healthy non-smoking females (n = 74) and males (n = 51) were studied with 22 commonly used lung function tests. If the data were logarithmically transformed, there was a satisfactory reduction of residual variation and the assumption of normality was fulfilled. Furthermore, the difference between males and females could be described by a binary sex-variable that is independent of interactions with other variables such as age, height and weight. Logarithmic transformation of lung function measurements followed by linear regression is recommended as a simple standard method for calculation of reference values.
如今可用的强大计算机能够通过回归模型的无穷排列来计算参考值，而这些模型并不总是具有生物学意义。本研究的目的是测试简单回归模型描述为计算参考值而收集的数据的能力。对74名健康非吸烟女性和51名健康非吸烟男性进行了22项常用肺功能测试。如果对数据进行对数转换，残差变异会得到令人满意的减少，并且满足正态性假设。此外，男性和女性之间的差异可以用一个二元性别变量来描述，该变量独立于与年龄、身高和体重等其他变量的相互作用。建议对肺功能测量值进行对数转换，然后进行线性回归，作为计算参考值的简单标准方法。
