Least square analysis

In multivariate least squares analysis, the dependent variable is a function of two or more independent variables. Because matrices are so conveniently handled by computer and because the mathematical formalism is simpler, multivariate analysis will be developed as a topic in matrix algebra rather than conventional algebra. 

We have already seen the normal equations in matrix form. In the multivariate case, there are as many slope parameters as there are independent variables and there is one intercept. The simplest multivariate problem is that in which there are only two independent variables and the intercept is zero 

To simplify the algebra, the error in the a variable will be considered negligible relative to the error in the y variable, although this is not a necessar y condition. Equation (3-54) describes a plane passing through the origin. Let us restrict it to positive values of Xi, a 2, and y. 

One measurement of the dependent variable yields yi for known values of the independent variables xu,xn 

The mathematical requirements for unique determination of the two slopes mi and ni2 are satisfied by these two measurements, provided that the second equation is not a linear combination of the first. In practice, however, because of experimental error, this is a minimum requirement and may be expected to yield the least reliable solution set for the system, just as establishing the slope of a straight line through the origin by one experimental point may be expected to yield the least reliable slope, inferior in this respect to the slope obtained from 2, 3, or p experimental points. In univariate problems, accepted practice dictates that we 

The goal is to determine a functional form for (a, b,. .., T) that can be used to design reactors. The simplest case is to suppose that the reaction rate has been measured at various values a,b. T. A CSTR can be used for these measurements as discussed in Section 7.1.2. Suppose J data points have been measured. The jXh point in the data is denoted as S/t-data aj,bj. Tj) where Uj, bj. 7 are experimentally observed values. Corresponding to this measured reaction rate will be a predicted rate, modeii p bj,7 ). The predicted rate depends on the parameters of the model e.g., on k,m,n,r,s. in Equation (7.4) and these parameters are chosen to obtain the best fit of the experimental 

The first equation shows that the data and model predictions are compared at the same values of the (nominally) independent variables. The second equation explicitly shows that the sum-of-squares depends on the parameters in the model. 

When kinetic measurements are made in batch or piston flow reactors, the reaction rate is not determined directly. Instead, an integral of the rate is measured, and the rate itself must be inferred. The general approach is as follows  

Conduct kinetic experiments and measure some response of the system, such as Oout- Call this data.  

Pick a rate expression and assume values for its parameters. Solve the reactor design equations to predict the response. Call this prediction.  

---

Henderson, G. Lecture Graphic Aids for Least-Squares Analysis, /. Chem. Educ. 1988, 65, 1001-1003. 

Values for fQi and K 2 for acids of the form H2A are determined from a least-squares analysis of data from a potentiometric titration. 

The generalized inverse method represents another formulation of multilinear least-squares analysis. All the usual assumptions involved with least squares apply. 

The pyrimidine ring is virtually flat. Its corrected bond lengths, as determined by a least-squares analysis of the crystal structure data for a unit cell of four molecules, are shown in formula (2) (60AX80), and the bond angles derived from these data show good agreement with those (3) derived by other means (63JCS5893) for comparison, each bond... 

If the reaetion rate depends on more than one speeies, use the method of exeess eoupled either with the half-life method or the differential method. If the method of exeess is not suitable, an initial rate plot may be eonstrueted by varying the eoneentration of one reaetant while the eoneentrations of the others are held eonstant. This proeess is repeated until the orders of reaetion of eaeh speeies and the speeifie reaetion rate are evaluated. At level 5, the least-squares analysis ean be employed. 

A weighted least-squares analysis is used for a better estimate of rate law parameters where the variance is not constant throughout the range of measured variables. If the error in measurement is corrected, then the relative error in the dependent variable will increase as the independent variable increases or decreases. 

The weighted least-squares analysis is important for estimating parameter involving exponents. Examples are the eoneentration time data... 

Therefore in applying weighted least-squares analysis to Eq. (2-83), each c = In c should be weighted inversely to o /c rather than to cr. ... 

It can be argued that the main advantage of least-squares analysis is not that it provides the best fit to the data, but rather that it provides estimates of the uncertainties of the parameters. Here we sketch the basis of the method by which variances of the parameters are obtained. This is an abbreviated treatment following Bennett and Franklin.We use the normal equations (2-73) as an example. Equation (2-73a) is solved for <2o-... 

Correct weighting procedures for least-squares analysis have been discussed in... 

Thus, a can be calculated (it is sometimes negligible), and the rate constants are evaluated graphically or by least-squares analysis the estimates of k and k must be consistent with the known stability constant. 

Table 6-2. Least-Squares Analysis of Kinetic Data in Table 6-1...

Show that if the relative error ajk is constant, then an unweighted linear Arrhenius least-squares analysis is correct. 

This reaction was studied spectrophotometrically by monitoring the absorbance at 830 nm, where Pu02+ absorbs. The paired values of time and absorbance are presented for one experiment in Table 2-4. Figure 2-5 shows the data treatment according to Eq. (2-35). Nonlinear least-squares analysis gives k = (9.49 0.22) X 102 L mol-1 s-1 and a calculated end point absorbance of 0.025 0.003. 

Better yet, a least-squares analysis of k versus [B]av is carried out. The order with respect to [A], the limiting reagent, is established from the fit of the data to a chosen rate law. Experiments over a range of [A]o are a preferable way to show the order in LA], At constant [B], will be the same regardless of [A]o if the rate is first-order with respect to [A],... 

This expression constitutes an improvement. There are two advantages. First, the statistical reliability of the data analysis improves, because the variance in [A] is about constant during the experiment, whereas that of the quantity on the left side of Eq. (3-27) is not. Proper least-squares analysis requires nearly constant variance of the dependent variable. Second, one cannot as readily appreciate what the quantity on the left of Eq. (3-27) represents, as one can do with [A]t. Any discrepancy can more easily be spotted and interpreted in a display of (A] itself. 

Lactones, nomenclature, 105 Linked-Atom Least-Squares analysis, 319 Lithium gellan, 386 Luteic acid, 8... 

This book, and many standard texts, emphasizes graphical techniques for htting data. These methods give valuable qualitative insights that may be missed with too much reliance on least-squares analysis. 

These equations contain only one unknown parameter, kiAj. Assume values for it and solve Equations (11.20) and (11.21) simultaneously. Compare the calculated results with the experimental measurements using nonlinear least-squares analysis as in Equation (7.8). This is the preferred, modern approach, but the precomputer literature relied on computationally simpler methods for fitting kiAj. 

L Stable and S. Wold, Partial least square analysis with cross-validation for the two-class problem a Monte Carlo study. J. Chemometrics, 1 (1987) 185-196. 

While principal components models are used mostly in an unsupervised or exploratory mode, models based on canonical variates are often applied in a supervisory way for the prediction of biological activities from chemical, physicochemical or other biological parameters. In this section we discuss briefly the methods of linear discriminant analysis (LDA) and canonical correlation analysis (CCA). Although there has been an early awareness of these methods in QSAR [7,50], they have not been widely accepted. More recently they have been superseded by the successful introduction of partial least squares analysis (PLS) in QSAR. Nevertheless, the early pattern recognition techniques have prepared the minds for the introduction of modem chemometric approaches. 

A drawback of the method is that highly correlating canonical variables may contribute little to the variance in the data. A similar remark has been made with respect to linear discriminant analysis. Furthermore, CCA does not possess a direction of prediction as it is symmetrical with respect to X and Y. For these reasons it is now replaced by two-block or multi-block partial least squares analysis (PLS), which bears some similarity with CCA without having its shortcomings. 

---