Outliers exclusion

Osterberg and Norinder [42] further analyzed a subset of the ATPase data published by Litman et al. [39]. The ATPase activity values were correlated by multivariate statistics with calculated descriptors (MolSurf descriptors) related to physicochemical properties suchaslipophilicity, polarity, polarizability and hydrogen bonding. After exclusion of one outlier and large molecules such as valinomycin, gramicidin S and so on, which were not handled by the MolSurf software, only 21 compounds were included in the study. Two models were derived model 1, based on... 

The standard deviation is caicuiated after exclusion of outlier or with robust statistics... 

The outlier result should only be used along with a preponderance of other scientifically valid evidence to provide additional support to invalidate an OOS result. It alone cannot identify the source of an extreme result and thus should not be used as the exclusive reason to invalidate. 

Inclusion or exclusion of outliers can significantly influence exposure assessments, and harmonized outlier rejection criteria are needed. The suitability of existing approaches (e.g. anything beyond three standard deviations is an outlier) to occupational exposure data sets, as well as the role of field study observations, needs to be considered. 

Deviating values identified as possible outliers cannot always be discarded automatically. Values should be included or excluded on a rational basis. Check the records of the dubious values and correct errors. In some cases, deviating values should be rejected because noncorrectable causes have been found, such as previously unrecognized conditions, quahfying individuals for exclusion from the group of reference individuals. 

Children s scores on the auditory comprehension scale of the Preschool Language Scale were also inversely associated with maternal-hair Hg concentrations (p = 0.0019). Scores declined approximately 2.5 points across the range of Hg concentrations. Additional analyses identified several outlier or influential data points, whose exclusion finm the analyses reduced the estimates of the Hg effect substantially, sometimes to nonsignificance. In the pilot phase of the SCDS, information was not collected on several key variables that frequently confoimd the association between neurotoxicant exposures and child development. Those variables are socioeconomic status, caregiver inteUigence, and qirality of the home environment. 

FIGURE 12.10. Relationships among Brazilian populations of Haematobia irritans based on principal components after the exclusion of the outliers. Reprinted with permission from Reference 63. 

Filtering refers to exclusion of bad probes on the array or the whole array. Visual inspection of array images and color-coded plot images are the first tools that should be used to detect outlier probes or arrays [23]. Clustering (described below) is another common way to identify outlier array among the whole set of arrays in a study. 

When outliers are observed, and they always occur when the limits of a QSAR model are exceeded, it is poor practice to omit them (as often indicated in the literature by Statistics are significantly improved upon exclusion of compound X ). An explanation for this deviation shall instead be attempted ... 

However, if the possible outlier xi = 15) is omitted, the median is barely changed to 5.00 while the mean x is shifted to 4.857 (Figure 8.3). Typically the median value is a more robust indicator of a typical value of a set of univariate data than is the mean, e.g., in comparisons of family incomes in two different countries if a small fraction of families can have very high incomes well removed from the vast majority this can lead to misleading conclusions based on the mean values. In analytical chemistry, the main value of comparisons exemplified by the fictional data in Figure 8.3 lies in their ability to highlight suspicious values x that should be examined as possible outliers whose exclusion can be justified by appropriate statistical tests (Section 8.2.7). 

Statistical Analysis of Results and Performance Characteristics. A common problem with most external quality assurance systems has been the determination of the true value for a control sample. Because the true value is an ideal concept, the data are evaluated against the estimated true value, which may be established from (1) the amount of analyte spiked to the test samples (2) a group of reference laboratories using definitive or reference methods (3) the consensus mean, which is the mean of the results obtained by the participants after exclusion of outlier. It is the responsibility of the coordinators to ensure and demonstrate that the reference values are reliable [6]. 

After exclusion of outlier according to Dixon, Grubb, and Cochran as presented in Ref. 5, the parameters (mean, standard deviation, etc.) are calculated. One-way analysis of variance and outlier tests are applied to estimate the components of variance, repeatability, and reproducibility parameters. In the evaluation report often several performance parameters are given, i.e., the arithmetic mean, geometric mean, consensus mean, mean of results from different groups of analyses, and so forth, and a histogram of all results at the very least. For the validation of each laboratory a performance score is calculated. 

With only 9 data points, the data distribute nicely about zero, and the magnitude does not (visually) appear to be a function of x. It can be concluded that the fit is adequate. Note that a single point with a very large residual is a candidate "outlier" and might be omitted if this can be justified (poor experimental procedure, other extenuating circumstances, etc.). Elowever, the arbitrary exclusion of outliers must be avoided. 

The scatter in the values after exclusion of outliers is 3-38 % of the mean, depending on the test. Apart from the very low scatter in the quick-stick method, the scatter is lower for methods with longer contact times. Prerequisites for such results are very uniformly defined test specimens and exact compliance with defined test conditions [27]. 

---