Least squares approximation

Application of Eq. (5.21) to a number of polymers gave the correct order of magnitude for AHm-AH(0). The equation cannot be used for an accurate prediction of AHm, however, because of lack of data for AH(0) and the approximate character of Eqs. (5.7) and (5.8). But Eq. (5.11) suggests that AHm will increase with increasing values of Cp (298) and of (Tm Tg). This is proved in Fig. 5.7, where the ratio AHm/Cp(298) is plotted against (Tm - Tg) for a number of polymers, for which values of AHm have been published. As a first approximation (least squares fit for the curve through the origin correlation coefficient = 0.98 and a standard deviation of about 25%) ... 

Except for 6, which was arbitrarily fixed at 6 = 1, and which was fixed by a procedure to be described below, all other parameters in eqns (13) and (14) have been numerically defined by an approximate least-squares fitting procedure such as to correct the excessive attractiveness of the potential surface when the DMBE is truncated at three-body terms. It is on this calibration procedure that we focus next. 

The standard least-squares approach provides an alternative to the Galerkin method in the development of finite element solution schemes for differential equations. However, it can also be shown to belong to the class of weighted residual techniques (Zienkiewicz and Morgan, 1983). In the least-squares finite element method the sum of the squares of the residuals, generated via the substitution of the unknown functions by finite element approximations, is formed and subsequently minimized to obtain the working equations of the scheme. The procedure can be illustrated by the following example, consider... 

This method, because it involves minimizing the sum of squares of the deviations xi — p, is called the method of least squares. We have encountered the principle before in our discussion of the most probable velocity of an individual particle (atom or molecule), given a Gaussian distr ibution of particle velocities. It is ver y powerful, and we shall use it in a number of different settings to obtain the best approximation to a data set of scalars (arithmetic mean), the best approximation to a straight line, and the best approximation to parabolic and higher-order data sets of two or more dimensions. 

The Least Squares or Best-fit Line. The simplest type of approximating curve is a straight line, the equation of which can be written as in form number 1 above. It is customary to employ the above definition when X is the independent variable and Y is the dependent variable. 

If the UCKRON expression is simplified to the form recommended for reactions controlled by adsorption of reactant, and if the original true coefficients are used, it results in about a 40% error. If the coefficients are selected by a least squares approach the approximation improves significantly, and the numerical values lose their theoretical significance. In conclusion, formalities of classical kinetics are useful to retain the basic character of kinetics, but the best fitting coefficients have no theoretical significance. 

For a function f(x) given only as discrete points, the measure of accuracy of the fit is a function d(x) = f(x) - g(x) where g(x) is the approximating function to f(x). If this is interpreted as minimizing d(x) over all x in the interval, one point in error can cause a major shift in the approximating function towards that point. The better method is the least squares curve fit, where d(x) is minimized if... 

The buildup of [P] is biexponential. Nonlinear least-squares fitting will provide the best solution for k and k2 (and, if they are unknown, for [A ]o and [A2]o also). Another method is to plot ln([P] - [P],) against time. At long times, the exponent with the higher value of k will have fallen off, such that the slope will approximate the slower one. Calling this ks (it might be either k or k2), we perform a subtraction, emphasizing the data at shorter times ... 

Toward these ends, the kinetics of a wider set of reaction schemes is presented in the text, to make the solutions available for convenient reference. The steady-state approach is covered more extensively, and the mathematics of other approximations ( improved steady-state and prior-equilibrium) is given and compared. Coverage of data analysis and curve fitting has been greatly expanded, with an emphasis on nonlinear least-squares regression. 

The isotherms can be approximated by the well-known method of least squares, in which the function F is to be minimized ... 

X-Ray diffraction from single crystals is the most direct and powerful experimental tool available to determine molecular structures and intermolecular interactions at atomic resolution. Monochromatic CuKa radiation of wavelength (X) 1.5418 A is commonly used to collect the X-ray intensities diffracted by the electrons in the crystal. The structure amplitudes, whose squares are the intensities of the reflections, coupled with their appropriate phases, are the basic ingredients to locate atomic positions. Because phases cannot be experimentally recorded, the phase problem has to be resolved by one of the well-known techniques the heavy-atom method, the direct method, anomalous dispersion, and isomorphous replacement.1 Once approximate phases of some strong reflections are obtained, the electron-density maps computed by Fourier summation, which requires both amplitudes and phases, lead to a partial solution of the crystal structure. Phases based on this initial structure can be used to include previously omitted reflections so that in a couple of trials, the entire structure is traced at a high resolution. Difference Fourier maps at this stage are helpful to locate ions and solvent molecules. Subsequent refinement of the crystal structure by well-known least-squares methods ensures reliable atomic coordinates and thermal parameters. 

The standard way to answer the above question would be to compute the probability distribution of the parameter and, from it, to compute, for example, the 95% confidence region on the parameter estimate obtained. We would, in other words, find a set of values h such that the probability that we are correct in asserting that the true value 0 of the parameter lies in 7e is 95%. If we assumed that the parameter estimates are at least approximately normally distributed around the true parameter value (which is asymptotically true in the case of least squares under some mild regularity assumptions), then it would be sufficient to know the parameter dispersion (variance-covariance matrix) in order to be able to compute approximate ellipsoidal confidence regions. 

The crystal structures of PbTX-1 dimethyl acetal, PbTX-1, and dihydro PbTX-1 provide a total of four independent pictures of the same brevetoxin skeleton. It is rare that this quantity of structural data is available for a natural product of this size. A comparison of torsional angles shows that all four molecules have approximately the same conformations in all rings, except, of course, for the aldehyde side chain and the E-ring in one of the independent molecules of PbTX-1. Least squares superposition fits among the four molecules gave the following average distances ... 

The selection to minimize absolute error [Eq. (6)] calls for optimization algorithms different from those of the standard least-squares problem. Both problems have simple and extensively documented solutions. A slight advantage of the LP solution is that it does not need to be solved for the points for which the approximation error is less than the selected error threshold. In contrast, the least squares problem has to be solved with every newly acquired piece of data. The LP problem can effectively be solved with the dual simplex algorithm, which allows the solution to proceed recursively with the gradual introduction of constraints corresponding to the new data points. 

Compression may be achieved if some regions of the time-frequency space in which the data are decomposed do not contain much information. The square of each wavelet coefficient is proportional to the least-squares error of approximation incurred by neglecting that coefficient in the reconstruction ... 

Fig. 2.39. Na /K+ atomic ratios of well discharges plotted at measured downhole temperatures. Curve A is the least squares fit of the data points above 80°C. Curve B is another emperical curve (from Truesdell, 1976). Curves C and D show the approximate locations of the low albite-microcline and high albite-sanidine lines derived from thermodynamic data (from Fournier, 1981). Small solid subaerial geothermal water Solid square Okinawa Jade Open square South Mariana Through Solid circle East Pacific Rise 11°N Open circle Mid Atlantic Ridge, TAG.

Y. Escoufier and S. Junca, Least squares approximation of frequencies or their logarithms. Int. Statistical Rev., 54 (1986) 279-283. 

An alternative and illuminating explanation of reduced rank regression is through a principal component analysis of Y, the set of fitted F-variables resulting from an unrestricted multivariate multiple regression. This interpretation reveals the two least-squares approximations involved projection (regression) of Y onto X, followed by a further projection (PCA) onto a lower dimensional subspace. 

Some of the results are collected in Table 35.7. Table 35.7a shows that some sensory attributes can be fitted rather well by the RRR model, especially yellow and green (/ == 0.75), whereas for instance brown and syrup do much worse R 0.40). These fits are based on the first two PCs of the least-squares fit (Y. The PCA on the OLS predictions showed the 2-dimensional approximation to be very good, accounting for 99.2% of the total variation of Y. The table shows the PC weights of the (fitted) sensory variables. Particularly the attributes brown , and to a lesser extent syrup , stand out as being different and being the main contributors to the second dimension. 

---