Regression multi-linear

This step involves calibration of the apparatus which will serve as a reference. It consists of analysing the greatest number possible (minimum 50) wines or must samples containing different and accurately known concentrations of each analyte. The concentration points should be uniformly distributed over the probable scale of measure for each analyte. The matrices should mimic as accurately as possible the wines and musts destined for analysis using that particular instrument. For each calibration sample, a measurement is carried out at a maximum number of wavelengths in the infra-red. Multi-linear regression is then carried out on the results which enables the following relationship to be established ... 

In the present study, two ANN methods - the FFBP with the Levenberg-Marquardt algorithm and the radial basis functions (RBF) - were employed to estimate the air pollution parameters measured at a station in Istanbul on chosen episode days, with the focus on the particulate matter. The results were compared with those obtained with the multi-linear regression (MLR) method. 

Another factorial design, used for studying solubility in mixed micelles, introduces and demonstrates multi-linear regression and analysis of variance. It is then extended, also in chapter 5, to a central composite design to illustrate the estimation of predictive models and their validation. 

The coefficients in the model equation 3.4 may be estimated as before, as linear combinations or contrasts of the experimental results, taking the columns of the effects matrix as described in section III.A.5 of chapter 2. Alternatively, they may be estimated by multi-linear regression (see chapter 4). The latter method is more usual, but in the case of factorial designs both methods are mathematically equivalent. 

As before, the coefficients may be estimated either by linear combinations (contrasts) corresponding to the 16 columns of the model matrix, or by multi-linear regression. These estimates are listed in table 3.9 and plotted in figure 3.8. 

Estimates of the statistical significance of the coefficients can and frequently should be obtained by other means - in particular by replicated experiments (usually centre points) followed by multi-linear regression of the data, and analysis of variance, as developed in chapter 4. The methods we have described above are complementary to those statistical methods and are especially useful for saturated designs of 12 to 16 or more experiments. For designs of only 8 experiments, the results of these analyses should be examined with caution. 

Another solution, by far the more general one, is to carry out the experiments in a random order. If there are the same number of experiments in the design as there are parameters in the postulated model, the effect of time is to perturb the different estimations in a random fashion. If the number of experiments exceeds the number of parameters to be estimated then the experimental error estimated by multi-linear regression (see chapter 4) includes the effect of time and is therefore overestimated with respect to its true value. 

We may take the results of the 2 factorial study of an effervescent table formulation reported earlier, and select the data corresponding to the 12 experiments of table 3.31. Estimates of the coefficients obtained either by contrasts or by the usual method of multi-linear regression are very close to those estimated from the... 

In the same way as for the % x 2 design, we can select the data for these 12 experiments from the experimental results of table 3.8 (2 effervescent tablet factor study) and estimate the coefficients by multi-linear regression. (The linear combinations method is not applicable here.) The results, given in table 3.36, are almost identical to those found with the full design and reported in table 3.9. 

Introduction to Multi-Linear Regression and Analysis of Variance... 

Multi-linear regression by the least squares method... 

C. Multi-Linear Regression by the Least Squares Method... 

Both the information and the dispersion matrices are of great importance, not only in determining coefficients by multi-linear regression, but also in accessing the quality of an experimental design. They will be referred to quite frequently in the remainder of this chapter and the remainder of the book. In our example the information and dispersion matrices are as follows ... 

Least squares multi-linear regression is by far the most common method for estimating "best values" of the coefficients, but it is not the only method, and is not always the best method. So-called "robust" regression methods exist and may be useful. These reduce the effect on the regression line of outliers, or apparently aberrant data points. They will not be discussed here, and least squares regression is used in the examples throughout this book. 

Derivation of the least squares multi-linear regression equation... 

The quadratic model parameters can be estimated by multi-linear regression and the model analysed by ANOVA. The adjusted model is ... 

The optimization of a gel formulation of ketoprofen has been described by Takayama and Nagai (4). They studied the effect of t/-limonene and ethanol in their formulation on the penetration rate of the drug through the skin of the rat in vivo. They also measured the lag time and the skin irritation. A central composite design with 4 centre points was used and predictive second-order equations were obtained by multi-linear regression. 

If the experimenter requires a more certain or more precise identification of the optimum, or to know how the response or responses vary about the final point, he should not try to obtain a mathematical model at the end of the optimization by multi-linear regression over the points (not even limiting the points to those about the optimum). The variance inflation factors are often very high, a measure of the lack of reliability of the regression. Rather, he should carry out a response surface study around the supposed optimum. 

Since in the latter case there are 3 degrees of freedom, the model coefficients may also be estimated by least squares multi-linear regression, instead of the direct method shown above. The values thus calculated are only slightly different from the previous estimates. 

Multi-linear regression and analysis of variance for mixture models... 

Determination of reduced cubic model by multi-linear regression... 

Instead of calculating the solubility at each of the test points and comparing it with the experimental value we fit the coefficients of the reduced cubic model to the data by least squares multi-linear regression, and investigate the goodness of fit by analysis of variance. The resulting equation is ... 

The reference-state coefficients are estimated by multi-linear regression. To return to the original coefficients (presence-absence model) we use the following relations ... 

The coefficients are calculated by multi-linear regression, according to the least squares method. There are a very large number of different programs for doing these calculations. The use of properly structured experimental designs, which are usually quite close to orthogonality, has the result that the more sophisticated methods (partial least squares etc.) are not usually necessary. 

Determination of model least squares multi-linear regression weighted multi-linear regression robust regression ridge regression partial least squares generalised inverse... 

---