论文部分内容阅读
同位素年龄测定是受从计数统计和实验误差中产生的不确定性支配的。当计算一个同位素年龄差时这些不确定性是附加上去的。如果这些不确定性很大,用“经典”统计方法会导致无意义的年龄差。在许多场合下,由于知道地层层序或其它线索,相对年龄往往是已知的。这种信息可用来建立年龄差的贝叶斯估计,它将把年龄顺序的先验知识包括进去。假定年龄测量误差服从对数正态分布,并对两个已知顺序的真实年龄采用了非情报性但却受约束的双变量的先验知识。真实年龄比是个截尾对数正态分布的变量。它的期望值给出一个年龄比的估计,它的方差提供了置信区间。对年龄的贝叶斯估计则与此不同,即使测量的年龄是相等的或是反序的,它也能给出正确的顺序。例如,对两个样品的年龄测量可能都得到100ka(千年),带有0.2的变化系数。对年龄差的贝叶斯估计则是22.7ka,带有75％置信区间为〔4.4,43.7〕ka。
Isotope age determination is governed by the uncertainty that arises from counting statistics and experimental errors. These uncertainties are additive when calculating the age difference of an isotope. If these uncertainties are large, using “classic” statistical methods can lead to meaningless age differences. On many occasions relative age is often known due to the sequence of stratigraphy or other clues. This information can be used to establish age-poor Bayesian estimates that will include prior knowledge of the age order. It is assumed that the age measurement error obeys a lognormal distribution and adopts a priori knowledge of univocal but constrained bivariate for the true age of two known orders. The true age ratio is a variable with a censored lognormal distribution. Its expected value gives an estimate of age ratio, and its variance provides confidence intervals. The Bayesian estimate of age differs from this, and it gives the correct order, even if the measured ages are equal or reversed. For example, both samples may have age measurements of 100 ka (millennium) with a coefficient of variation of 0.2. The Bayesian estimate of age is 22.7ka with a 75% confidence interval of [4.4, 43.7] ka.