MODELING OF STATISTICAL TABLES FOR OUTLIER TESTS

被引：4

作者：

BUGNER, E ^{[1
]}

RUTLEDGE, DN ^{[1
]}

机构：

[1] INST NATL AGRON PARIS GRIGNON,CHIM ANALYT LAB,F-75231 PARIS 05,FRANCE

来源：

CHEMOMETRICS AND INTELLIGENT LABORATORY SYSTEMS | 1990年 / 9卷 / 03期

关键词：

D O I：

10.1016/0169-7439(90)80076-I

中图分类号：

TP [自动化技术、计算机技术];

学科分类号：

0812 ;

摘要：

Non-linear fitting was applied to model the statistical tables used in two very common tests used to detect outliers: the Dixon and the Cochran tests. The Marquardt algorithm was used to find multiexponential approximations of the data. It was possible in this way to greatly reduce the number of values to be incorporated in software using these tests. The errors due to the models are in all cases smaller than the rounding errors of the tables. The original values can be easily recalculated using simple formulae which is far more convenient for a programmer than a table. This method also gives one the possibility of extrapolating the tables to a larger sample size.

引用

页码：257 / 259

页数：3