Data handling confidence

Hyphenated analytical methods usually give rise to iacreased confidence ia results, eaable the handling of more complex samples, improve detectioa limits, and minimi2e method development time. This approach normally results ia iacreased iastmmeatal complexity and cost, iacreased user sophisticatioa, and the need to handle enormous amounts of data. The analytical chemist must, however, remain cogni2ant of the need to use proper analytical procedures ia sample preparatioas to aid ia improved seasitivity and not rely solely on additional iastmmentation to iacrease detection levels. 

It is necessary to work with the manufacturer in sizing and rating these special units, because sufficient public data/ correlation of heat transfer does not exist to allow the design engineer to handle the final and detailed design with confidence. 

The instrument has been evaluated by Luster, Whitman, and Fauth (Ref 20). They selected atomized Al, AP and NGu as materials for study that would be representative of proplnt ingredients. They found that only 2000 particles could be counted in 2 hours, a time arbitrarily chosen as feasible for control work. This number is not considered sufficient, as 18,000 particles are required for a 95% confidence level. Statistical analysis of results obtained for AP was impossible because of discrepancies In the data resulting from crystal growth and particle agglomeration. The sample of NGu could not be handled by the instrument because it consisted of a mixt of needles and chunky particles. They concluded that for dimensionally stable materials such as Al or carborundum, excellent agreement was found with other methods such as the Micromerograph or visual microscopic count. But because of the properties peculiar to AP and NGu, the Flying Spot Particle Resolver was not believed suitable for process control of these materials... 

There are a number of other problems relating to the manipulation and interpretation of data that cause difficulty. The most common are (i) uncertainty about the number of replicate results required for proper comparison of the certified reference value, and (2) the actual analytical result and how gross outlier results should be handled. These issues and how to deal with data that falls outside the confidence limit are reviewed in detail by Walker and Lumley (1999), who conclude that whilst customer requirements may provide answers the judgement of the analyst must always be the final arbiter in any decision ... 

One disadvantage of the KNN method is that it does not provide an assessment of confidence in the class assignment result, hi addition, it does not sufficiently handle cases where an unknown sample belongs to none of the classes in the calibration data, or to more than one class. A practical disadvantage is that the user must input the number of nearest neighbors (K) to use in the classifier. In practice, the optimal value of K is influenced by the total number of calibration samples (N), the distribution of calibration samples between classes, and the degree of natural separation of the classes in the sample space. 

Providing checks on the laboratory data. For example, disagreement between electrical conductance measured in the field and TDI determined in the laboratory indicates erroneous field measurements, bad handling of the samples on the way to the laboratory, erroneous laboratory results, or mislabeling of samples. Agreement between field and laboratory data raises the confidence in the data and indicates that no secondary processes have occurred between sampling and laboratory measurement. 

Table 1.1 Effect of various factors on pesticide exposure. All data are unit exposure values ( xg/kg of active ingredient (a.i.) handled), taken from PHED (1992). Values are central tendency measures based on high confidence data sets...

The reliable, robust classification of diseases or disease states via biomedical spectroscopy requires special methodology that can handle complex data. In most cases, such data defy simple analyses that assume the presence of easily identified features ( markers ) in the data set. In particular, the methodology must be able to handle data sets that contain relatively few spectra (in the 50s) but many attributes (data points) per spectrum (in the 1000s) ideally, it should also provide some measure of the degree of confidence in a given diagnosis. 

Quality Control Data. Data obtained from assays of blood gas and pH control materials may be handled in the same way as data from other clinical chemistry determinations (i.e., mean, SD, and coefficient of variation, and control and confidence limits for construction of Levey-Jennings plots). As stability of commercial aqueous control materials is generally several months, vendors often provide data reduction programs that standardize and simplify documentation. However, the resulting reports are temporally delayed and are most useful for meeting accreditation requirements as opposed to real-time corrective or preventive action. They are however useful to compare long-term performances with other laboratories. Equally important features of quality assurance to an active blood gas service are the sixth sense of practiced operators for detecting subtle manifestations of deterioration of instrument performance and the suspicion of trouble expressed by clinicians. 

Summary We have seen how data from experimental studies of equilibria are converted into a readily usable form, that is, into AH, AS and AG values. These data are combined with others, determined, for example, by reaction calorimetry (AH) or by Third Law (as opposed to Second Law ) determinations of entropy, using low-temperature heat capacity measurements. Before we can enter the first division of predictive thermodynamics, and confidently plan processes, we must learn to accept information from two further sources. First, there is a great reservoir of electrochemical expertise which we have not yet tapped. Secondly, we must learn to handle the refinements in free energy formulations, which make allowance for the slight variations of AH and AS with temperature. 

The use of Weibull plots for design purposes has to be handled with extreme care. As with all extrapolations, a small uncertainty in the slope can result in large uncertainties in the survival probabilities, and hence to increase the confidence level, the data sample has to be sufficiently large N > 100). Furthermore, in the Weibull model, it is implicitly assumed that the material is homogeneous, with a single flaw population that does not change with time. It further assumes that only one failure mechanism is operative and that the defects are randomly distributed and are small relative to the specimen or component size. Needless to say, whenever any of these assumptions is invalid, Eq. (11.23) has to be modified. For instance, bimodal distributions that lead to strong deviations from a linear Weibull plot are not uncommon. 

---