Errors chemometrics

Procedures used vary from trial-and-error methods to more sophisticated approaches including the window diagram, the simplex method, the PRISMA method, chemometric method, or computer-assisted methods. Many of these procedures were originally developed for HPLC and were apphed to TLC with appropriate changes in methodology. In the majority of the procedures, a set of solvents is selected as components of the mobile phase and one of the mentioned procedures is then used to optimize their relative proportions. Chemometric methods make possible to choose the minimum number of chromatographic systems needed to perform the best separation. 

Another recent tool has been developed within the ORCHESTRA project. The tool keeps into account both the chemometric information and the toxicity predictions done by the model, and in particular what kind of errors have been done by the model. It applies to the CAESAR QSAR models. Furthermore, this tool is based not only on the a priori data and information, as the other approaches, but also on the a posteriori result of the model. The user knows if the model can or cannot be used for a certain compound. In some cases a warning is given, recommending expert opinion. In all cases the reasons for the reliability is given, and it can be evaluated in a transparent way. 

Lorber A, Kowalski BR (1988) Estimation of prediction error for multivariate calibration. J Chemometrics 2 93... 

In both experiments, Conditions 1 and 2 together mean that all results from the experiment will be the same in the first scenario, and all results except the ones corresponding to the effective catalyst will be the same while that one will differ. Condition 3 means that we do not need to use any statistical or chemometric considerations to help explain the results. However, for pedagogical purposes we will examine this experiment as though random error were present, in order to be able to compare the analyses we obtain in the presence and in the absence of random effects. The data from these two scenarios might look like that shown in Table 10-4. 

This leads us to the other hand, which, it should be obvious, is that we feel that Chemometrics should be considered a subfield of Statistics, for the reasons given above. Questions currently plaguing us, such as How many MLR/PCA/PLS factors should I use in my model , Can I transfer my calibration model (or more importantly and fundamentally How can I tell if I can transfer my calibration model ), may never be answered in a completely rigorous and satisfactory fashion, but certainly improvements in the current state of knowledge should be attainable, with attendant improvements in the answers to such questions. New questions may arise which only fundamental statistical/probabilistic considerations may answer one that has recently come to our attention is, What is the best way to create a qualitative (i.e., identification) model, if there may be errors in the classifications of the samples used for training the algorithm ... 

One part of that equation, [AtA] , appears so commonly in chemometric equations that it has been given a special name, it is called the pseudoinverse of the matrix A. The uninverted term ATA is itself fairly commonly found, as well. The pseudoinverse appears as a common component of chemometric equations because it confers the Least Squares property on the results of the computations that is, for whatever is being modeled, the computations defined by equation 69-1 produce a set of coefficients that give the smallest possible sum of the squares of the errors, compared to any other possible linear model. 

Sections on matrix algebra, analytic geometry, experimental design, instrument and system calibration, noise, derivatives and their use in data analysis, linearity and nonlinearity are described. Collaborative laboratory studies, using ANOVA, testing for systematic error, ranking tests for collaborative studies, and efficient comparison of two analytical methods are included. Discussion on topics such as the limitations in analytical accuracy and brief introductions to the statistics of spectral searches and the chemometrics of imaging spectroscopy are included. 

A critical attitude towards the results obtained in analysis is necessary in order to appreciate their meaning and limitations. Precision is dependent on the practical method and beyond a certain degree cannot be improved. Inevitably there must be a compromise between the reliability of the results obtained and the use of the analyst s time. To reach this compromise requires an assessment of the nature and origins of errors in measurements relevant statistical tests may be applied in the appraisal of the results. With the development of microcomputers and their ready availability, access to complex statistical methods has been provided. These complex methods of data handling and analysis have become known collectively as chemometrics. 

Nonlinear mapping (NLM) as described by Sammon (1969) and others (Sharaf et al. 1986) has been popular in chemometrics. Aim of NLM is a two-(eventually a one- or three-) dimensional scatter plot with a point for each of the n objects preserving optimally the relative distances in the high-dimensional variable space. Starting point is a distance matrix for the m-dimensional space applying the Euclidean distance or any other monotonic distance measure this matrix contains the distances of all pairs of objects, due. A two-dimensional representation requires two map coordinates for each object in total 2n numbers have to be determined. The starting map coordinates can be chosen randomly or can be, for instance, PC A scores. The distances in the map are denoted by d t. A mapping error ( stress, loss function) NLm can be defined as... 

All regression methods aim at the minimization of residuals, for instance minimization of the sum of the squared residuals. It is essential to focus on minimal prediction errors for new cases—the test set—but not (only) for the calibration set from which the model has been created. It is relatively easy to create a model— especially with many variables and eventually nonlinear features—that very well fits the calibration data however, it may be useless for new cases. This effect of overfitting is a crucial topic in model creation. Definition of appropriate criteria for the performance of regression models is not trivial. About a dozen different criteria— sometimes under different names—are used in chemometrics, and some others are waiting in the statistical literature for being detected by chemometricians a basic treatment of the criteria and the methods how to estimate them is given in Section 4.2. 

Determination of the optimum complexity of a model is an important but not always an easy task, because the minimum of measures for the prediction error for test sets is often not well marked. In chemometrics, the complexity is typically controlled by the number of PLS or PCA components, and the optimum complexity is estimated by CV (Section 4.2.5). Several strategies are applied to determine a reasonable optimum complexity from the prediction errors which may have been obtained by CV (Figure 4.4). CV or bootstrap allows an estimation of the prediction error for each object of the calibration set at each considered model complexity. 

Not just by accident PLS regression is the most used method for multivariate calibration in chemometrics. So, we recommend to start with PLS for single y-variables, using all x-variables, applying CV (leave-one-out for a small number of objects, say for n < 30, 3-7 segments otherwise). The SEPCV (standard deviation of prediction errors obtained from CV) gives a first idea about the relationship between the used x-variables and the modeled y, and hints how to proceed. Great effort should be applied for a reasonable estimation of the prediction performance of calibration models. 

Crowley et al. performed some interesting work with Pichia pastoris in a fed-batch process.19 The complex mixture was measured using a multibounce attenuated total reflectance (HATR) cell. The authors developed models for glycerol, methanol, and the product, a heterologous protein. The results are reported somewhat differently from normal chemometric results. The authors used root-mean square error (RMSE) for the product as a performance index and measured a percent difference for the methanol and glycerol. 

The calibration problem in chromatography and spectroscopy has been resolved over the years with varying success by a wide variety of methods. Calibration graphs have been drawn by hand, by instruments, and by commonly used statistical methods. Each method can be quite accurate when properly used. However, only a few papers, for example ( 1,2,15,16,26 ), show the sophisticated use of a chemometric method that contains high precision regression with total assessment of error. 

However, society likes to have decisions made in a black and white manner and to know whether something is there or not. This situation suggests that the analytical error should drop to zero. While this result is the goal of all analytical work, it is simply not realistic. Our basic need, then, is to simplify error determinations and explanations and to educate the public both for the reasons and for the interpretations of error. The goal of this volume is to further the use of mathematical and statistical tools—the field of chemometrics—for chemical and, specifically, trace chemical analyses of pesticides and environmental contaminants. 

Statistics have been used in chemical analysis in increasing amounts to quantify errors. The focus shifts now to other areas, such as in sampling and in measurement calibrations. Statistical and computer methods can be brought into use to give a quantified amount of error and to clarify complex mixture problems. These areas are a part of chemometrics as we use the term today. 

This chapter can be viewed as a massive refutation, complete with all alternative solutions, of the fatal flaw in the last sentence above. Chemometrics most emphatically does not eliminate sampling errors [sic]. 

Figure 3.10 Hallmark signature of significant sampling bias as revealed in chemometric multivariate calibrations (shown here as a prediction validation). Crab sampling results in an unacceptably high, irreducible RMSEP. While traditionally ascribed to measurement errors, it is overwhelmingly due to ISE.

K.H. Esbensen, H.H. Eriis-Petersen, L. Petersen, J.B. Hohn-Nielsen and P.P. Mortensen, Representative process sampling - in practice variographic analysis and estimation of Total Sampling Errors (TSE). Proceedings 5 th Winter Symposium of Chemometrics (WSC-5), Samara 2006. Chemom. Intell. Lab. Syst, Special Issue, 88(1), 41—19 (2007). 

The f-test is similar to the t-test, but is used to determine whether two different standard deviations are statistically different. In the context of chemometrics, the f-test is often used to compare distributions in regression model errors in order to assess whether one model is significantly different than another. The f-statistic is simply the ratio of the squares of two standard deviations obtained from two different distributions ... 

Chemometrics in Process Analytical Technology (PAT) 409 The main figure of merit in test set validation is the root mean square error of prediction (RMSEP) ... 

Outliers demand special attention in chemometrics for several different reasons. During model development, their extremeness often gives them an unduly high influence in the calculation of the calibration model. Therefore, if they represent erroneous readings, then they will add disproportionately more error to the calibration model. Furthermore, even if they represent informative information, it might be determined that this specific information is irrelevant to the problem. Outliers are also very important during model deployment, because they can be informative indicators of specific failures or abnormalities in the process being sampled, or in the measurement system itself. This use of outlier detection is discussed in the Model Deployment section (12.10), later in this chapter. 

Frake et al. compared various chemometric approaches to the determination of the median particle size in lactose monohydrate with calibration models constrncted by MLR, PLS, PCR or ANNs. Overall, the ensuing models allowed mean particle sizes over the range 20-110/tm to be determined with an error less than 5 pm, which is comparable to that of the laser light diffraction method nsed as reference. Predictive ability was similar for models based on absorbance and second-derivative spectra this confirms that spectral treatments do not suppress the scattering component arising from differences in particle size. 

In this chapter, three chemometric methods of increasing importance to SEC are examined nonlinear regression, graphics and error propagation analysis. These three methods are briefly described with emphasis on SEC applications and on critical concerns in their correct implementation. In addition to the specific references cited, further information on these methods and others may be found in a recent book vdiich examines chemometrics in both SEC and HPIiC together (J ) as well as in periodic reviews (2) ... 

---