Test data validation

A9.3.6.2.4 In the absence of empirical test data, validated Quantitative Structure Activity Relationships... 

Rectification accounts for systematic measurement error. During rectification, measurements that are systematically in error are identified and discarded. Rectification can be done either cyclically or simultaneously with reconciliation, and either intuitively or algorithmically. Simple methods such as data validation and complicated methods using various statistical tests can be used to identify the presence of large systematic (gross) errors in the measurements. Coupled with successive elimination and addition, the measurements with the errors can be identified and discarded. No method is completely reliable. Plant-performance analysts must recognize that rectification is approximate, at best. Frequently, systematic errors go unnoticed, and some bias is likely in the adjusted measurements. 

Tests data against valid ranges (e.g.. pH<]4) or lists of acceptable data (e.g., chemical names). 

Suppose for example that we want to test the validity of our data against the (larraonic model... 

Divide the available data into training and test data sets (1/3). Test sets are used to validate the trained network and insure accurate generalization. 

Graph the above data in the form Cp,m/T against T2 to test the validity of the Debye low-temperature heat capacity relationship [equation (4.4)] and find a value for the constant in the equation, (b) The heat capacity study also revealed that quinoline undergoes equilibrium phase transitions, with enthalpies as follows ... 

The recipe (5.58) is even more sensitive to the high-frequency dependence of kjj than similar criterion (5.53), which was used before averaging over kinetic energy of collisions E. It is a much better test for validity of microscopic rate constant calculation than the line width s j-dependence, which was checked in Fig. 5.6. Comparison of experimental and theoretical data on ZR for the Ar-N2 system presented in [191] is shown in Fig. 5.7. The maximum value Zr = 22 corresponding to point 3 at 300 K is determined from the rate constants obtained in [220],... 

One goal of tropospheric [HO ] or [H02 ] measurements is the generation of data for comparison with model calculations-to test or validate the models. Due to its high reactivity, HO comes into rapid photochemical equilibrium with its surroundings. Thus a test of a photochemical model, which compares measured and calculated HO concentrations, is mainly a test of the chemical mechanism that the model contains, and is relatively independent of... 

When testing and analysis are completed, the data can be analyzed and summarized. Statistical methods are often used during this step In a study. Data should first be edited and validated. Quality assurance Information from both the sampling and laboratory analyses should be considered In this validation. Field sampling personnel and laboratory scientists should maintain responsibility for data validation. 

QA SQPs should specify the amount of data to be audited and how the data points are chosen for audit. An auditor may choose to perform more thorough and more frequent audits on a recently validated system. The validation report can be used to assist in determining what and how much to audit. For example, if data summary printouts from the chromatographic computer system are used in the report, the validation report should be reviewed to verify that this summary function was tested during validation. If this portion of the computer software was successfully validated, verifying a few values from each table in the report may be sufficient. 

Independent studies (Cybenko, 1988 Homik et al., 1989) have proven that a three-layered back propagation network will exist that can implement any arbitrarily complex real-valued mapping. The issue is determining the number of nodes in the three-layer network to produce a mapping with a specified accuracy. In practice, the number of nodes in the hidden layer are determined empirically by cross-validation with testing data. 

Thus, the process of model testing and validation (considered synonymous) should ideally include the steps of calibration (if necessary), verification, and post-audit analyses. I indicate "ideally" because in many applications existing data will not support performance of all steps. In chemical fate modeling, chemical data for verification is often lacking and post-audit analyses are rare (unfortunately) for any type of modeling exercise. 

We have applied the methods described above to a number of data sets involving chemical reactivities, chemical properties and physical properties. In several cases the data sets were chosen because they provided an opportunity to test the validity of some of the parameters estimated in this work. All of the data sets studied are reported in Table 9. 

The d-d spectra of copper(II) compounds have provided a fruitful field for practitioners of the AOM. A wealth of structural data is available, and a rich variety of coordination geometries has been revealed. If we can make allowance for the dependence of the AOM parameters on the intemuclear distance, we are provided with excellent opportunities to test the validity of AOM parameters over a range of related systems. However, the progress of such studies over the years has illustrated the fact that a simple model may be very successful in explaining a limited amount of dubious experimental data as more crystal structures appear and as better spectroscopic data become available, the simple model may require considerable refurbishment, perhaps to the extent that it loses some of its appeal and utility. 

To test the validity of the extended Pitzer equation, correlations of vapor-liquid equilibrium data were carried out for three systems. Since the extended Pitzer equation reduces to the Pitzer equation for aqueous strong electrolyte systems, and is consistent with the Setschenow equation for molecular non-electrolytes in aqueous electrolyte systems, the main interest here is aqueous systems with weak electrolytes or partially dissociated electrolytes. The three systems considered are the hydrochloric acid aqueous solution at 298.15°K and concentrations up to 18 molal the NH3-CO2 aqueous solution at 293.15°K and the K2CO3-CO2 aqueous solution of the Hot Carbonate Process. In each case, the chemical equilibrium between all species has been taken into account directly as liquid phase constraints. Significant parameters in the model for each system were identified by a preliminary order of magnitude analysis and adjusted in the vapor-liquid equilibrium data correlation. Detailed discusions and values of physical constants, such as Henry s constants and chemical equilibrium constants, are given in Chen et al. (11). 

Cross validation and bootstrap techniques can be applied for a statistically based estimation of the optimum number of PCA components. The idea is to randomly split the data into training and test data. PCA is then applied to the training data and the observations from the test data are reconstmcted using 1 to m PCs. The prediction error to the real test data can be computed. Repeating this procedure many times indicates the distribution of the prediction errors when using 1 to m components, which then allows deciding on the optimal number of components. For more details see Section 3.7.1. 

---