Uncertainty in data integration

Uncertainty in Data Integration and Dataspace Support Platforms... 

This chapter is largely based on previous papers (Dong et al. 2007 Sarma et al. 2008). The proofs of the theorems we state and the experimental results validating some of our claims can be found there in. We also place several other works on uncertainty in data integration in the context of the system we envision. In the next section, we describe an architecture for data integration system that incorporates uncertainty. 

Although a torsion test is simple to carry out, it is not commonly accepted as an integral part of a material specification furthermore, few torsion data exist in handbooks. If, as is usually the case, the design needs to be based on tensile data, then a criterion of elastic failure has to be invoked, and this introduces some uncertainty in the calculated yield pressure (8). 

To provide a practical, understandable and common way of measurement uncertainty calculations, mainly based on already existing quality control and validation data covering all uncertainty sources in a integral way... 

The products of these transformations are pseudo enantiomers differing in absolute configuration and in mass, integration of the MS peaks and data processing affording the ee or E values. Any type of ionization can be employed, but electrospray ionization (ESI) is used most commonly [20,33-35]. An internal standard is advisable if it is necessary to determine percent conversion. The uncertainty in the ee value is less than 5%. In the original version about 1000 ee values could be measured per day [20a], but this has recently been increased to about 10 000 sam-... 

It is noteworthy that the styrene reference concentration (RfC) in the Integrated Risk Information System is based on the biomarker-response relationship found in workers (Mutti et al. 1984 EPA 1998). The Environmental Protection Agency (EPA) used the relationship of urinary biomarker to ambient-air concentration of workers to develop an RfC that was adjusted for the difference in exposure time between the workplace and the general population. That is a valid approach because it derives a workplace concentration-toxicity relationship in workers, which can then be adjusted for the general population to account for differences in exposure time and can take uncertainty factors into account. It is different from direct adjustment of the styrene BEI to evaluate human population biomonitoring data on styrene metabolites in urine, which would have the uncertainties described above and in Chapter 5. 

The negative peak in the baseline at 4520 cm-1 proved to be a convenient reference position. Its origin is presently unknown, but it likely arises from slight differences between the silica DIT cell used for the sample and that used for the carbon tetrachloride reference. Band integration did not work veil for quantitation in this study, probably because of uncertainties in the data above 5264 cm-1 where the discontinuity due to the electronic filter in the spectrometer occurs. 

There are a number of factors which affect the emission rates of biogenic sulfur from wetlands. In a recent study these have been investigated for wetlands in Florida, USA, 157-59) and are summarized in a chapter in this volume (60). These factors are divided into spatial, seasonal, diel and tidal components. In addition, other variables which affect emissions are temperature, insolation, and soil inundation. When these factors are taken into account in estimating emissions, and emission rates are obtained by integrating over the appropriate cycle, the emission estimates are up to two orders of magnitude lower than earlier estimates. However, using these methods results in large uncertainties in the emission estimates, and considerable additional data are required to better refine and extend emission estimates to other environments. 

The ability to make a judgement regarding representativeness of data is closely related to data quality issues. For example, representativeness is closely associated with the data quality issue of appropriateness. The data quality issues of accuracy and integrity are closely related to quantification or characterization of variability and uncertainty. The data quality issue of transparency pertains to clear documentation that, in turn, can provide enough information for an analyst to judge the representativeness and characterize variability and uncertainty. 

All four hallmarks of quality—appropriateness, accuracy, integrity and transparency—also reflect factors that can be important contributors to the understanding and characterization of uncertainty in a predicted exposure. What kinds of biases or uncertainties might use of particular data contribute to overall uncertainty How much uncertainty do they contribute The quality of data to a large degree determines what approaches, whether qualitative and quantitative, can be taken to characterize uncertainty (see Part 1 of this document). 

---