Data performance problems

Several case studies concerning baseline and data performance problems, as well as troubleshooting examples, are selected here to illustrate problem diagnosis and resolution. 

---

Used either as prelaboratory preparation for related laboratory activities or to expose students to additional laboratory activities not available in their program, these modules motivate students to learn by proposing real-life problems in a virtual environment. Students make decisions on experimental design, observe reactions, record data, interpret these data, perform calculations, and draw conclusions from their results. Following a summary of the module, students test their understanding by applying what they have learned to new situations or by analyzing the effect of experimental errors. 

By definition, the exemplar patterns used by these algorithms must be representative of the various pattern classes. Performance is tied directly to the choice and distribution of these exemplar patterns. In light of the high dimensionality of the process data interpretation problem, these approaches leave in question how reasonable it is to accurately partition a space such as R6+ (six-dimensional representation space) using a finite set of pattern exemplars. This degradation of interpretation performance as the number of possible labels (classes) increases is an issue of output dimensionality. 

Time and space integrated, paired-data performance. Spacially and/or temporally integrated data can be compared to analogous model predictions, such as daily or monthly averages or totals. This can circumvent some of the timing problems noted in (J ). 

But for ordinary data, we would not expect such a sequence to happen. This is the reason most statistics work as general indicators of data performance the special cases that cause them to fail are themselves low-probability occurrences. In this case the problem is not whether or not the data are nonlinear, the problem is that they are nonrandom. This is a perfect example of the data failing to meet a criterion other than the one you are concerned with. Therefore the Durbin-Watson test fails, as would any statistical test fail for such data they are simply not amenable to meaningful statistical calculations. Nevertheless, a blind computation of the Durbin-Watson statistic would give an apparently satisfactory value. But this is a warning that other characteristics of the data can cause it to appear to meet the criteria. 

In this chapter we have stated the data reconciliation problem and explored some available techniques for solving it. It must be noted that the performance of reconciliation methods is strongly dependent on the particular system. 

Diagnostic approaches typically involve collecting data from a variety of sources. In addition to fact-based analyses, such as competitive benchmarking or value driver analyses, surveys and interviews can help in gaining a better understanding of the company s workings and uncovering the complex roots of performance problems ... 

Personnel Bias Because of the possibility of several interpretations of the plant-performance problem, the judgment of analysts plays a critical role. Any bias in the analysts judgments will carry through the data analyses. To minimize this, analysts must develop an... 

The Quality Assurance organization has no mandated role in the development of computer systems other than defining QA fnnctional reqnirements. Once the system has been validated, accepted, and installed, QA will be responsible for monitoring data collection until its reliability is confirmed in accordance with SOP, compliance of nser SOPs, training, and secnrity policies. Any performance problems shonld be commnnicated to the responsible management persotmel in a timely fashion. QA shonld monitor corrective actions and nnschednled downtime records. 

Although the UCR EPA chamber has only been in operation for a relatively short time, it has already obtained usefiil information concerning the performance of current mechanism in predicting the effects of VOCs and NOx on ozone formation. As discussed here and also in our companion paper (Carter, 2004a), the SAPRC-99 mechanism predicts O3 formation reasonably well in low NOx experiments, and in ambient simulation experiments at the high ROG/NOx levels where maximum ozone formation potentials are achieved. However, die new data indicate problems with the mechanisms that were not previously realized. The SAPRC-99 mechanism consistently under predicts O3 formation in the lower ROG/NOx experiments where O3 formation is most sensitive to VOCs, and the problem is even worse for CB4. Other experiments indicate that there are problems with the formulation wifli die current aromatics photooxidation mechanisms. It is possible that the problems with the under prediction at low ROG/NOx ratios may be caused by problems with the aromatics mechanisms. Experimental and mechanism development work to investigate and hopefully resolve these problems is underway... 

In typical data mining problems, very little is known in advance about the functional form of the mapping between X and Y. The most important factor in favoring one certain functional form over the other is by comparing the predictive performance of the derived models. The performance of a regression model is primarily judged based on the difference between the actual and predicted values for... 

Most systems, especially for use in industry, include an autosampler that allows unattended, overnight operation of the instrument. Computer software in commercial instruments controls the autosampler and the instrument, collects the data, performs the calculations, prints out the results, and shuts down the instrument when the analysis is completed. Most instruments have computer-controlled safety interlocks that shut down the plasma or the instrument and autosampler in the event a problem is detected. 

Before discussing more complicated models and systems with more than three components, we present a mixture design problem with real data, performed in the laboratory of Professor G. Oliveira Neto (Chemistry Department, Campinas State University). 

Performance problems with complex model data will likely be handled by intelligent software algorithms (occlusion culling, motion LODs) in the near future. 

---