Scaling the Data

It is also possible to scale the raw data prior to performing PCA. 

Each successive row in a data matrix formed from a coupled chromatogram corresponds to a spectrum taken at a given elution time. One of the simplest methods of scaling 

Scores of dataset B normalised over the first two principal components (a) Over entire region (b) Expansion of central regions 

Note that some people call this normalisation, but we will avoid that terminology, as this method is distinct from that in Section 6.2.2. The influence on PC scores plots has already been introduced (Chapter 4, Section 4.3.6.2) but will be examined in more detail in this chapter. 

Scores plot of dataset A, each row summed to a constant total, PC2 versus PCI 

---

Note that there are again different options for scaling the data. The variables could be scaled using the empirical standard deviations of the variables, or by using robust versions (see Section 2.2.2). The latter option should be preferred if outliers are present or if the data are inhomogeneous (for instance, divided into groups). 

If the original data contain Information on the uncertainties associated with each measurement the sensitivity of the variance of the results to these errors can be studied. Approaches Include uncertainty weighting during the autoscaling procedure which Is provided for In ARTHUR, uncertainty scaling (the data standard deviation used for autoscaling Is replaced by the measurement absolute error such as presented In Table VII), and Monte Carlo simulation for estimating the variance of the statistics based on the error perturbed data ( ). 

The activity of In has been determined electrochemically at 627, 700, and 800°C (Hoshino et al, 1965 Chatterji and Smith, 1973 Anderson, 1978). The fits are listed in the first three columns of Table III and the experimental points and calculated curves for 627°C are shown in Fig. 2. The enthalpies of mixing in In-Sb liquid have been determined for various temperatures between 627 (Hultgren et al, 1973b) and 911°C. The data of Predel and Oehme (1976) at 680°C are tabulated. For the other temperatures we have scaled the data (Predel, 1979) from a graph. Figures 3 and 4 show the calculated curves and... 

Both PCA and PLS are sensitive to the scaling of the variables. It is therefore customary to scale the data variable-wise to zero mean and unit variance. However, the possibility to employ a different scaling whenever this seems to be appropriate should be kept in mind. For instance, a well-known important variable may be given a larger variance, for example, 3.0 instead of 1.0. Another situation in which a different scaling may be considered is when the variables are obviously blocked in such a way that each block contains closely related variables. Each block may then be given equal variance whereby the blocks have equal chances to influence the direction of the PC or PLS dimension. 

It is possible to improve the method by scaling the data, but it is important to be very careful to think about the consequences of the various methods employed. It is sometimes possible to scale first the two way data and then unfold. However, a final centring should normally be performed on the unfolded matrix. In addition, variable selection can have a significant influence on the effectiveness of unfolded PLS models, since not all the 2040 variables are going to be particularly relevant or informative. 

Note that there are two conformers, one with the two H8 atoms on the same side of the coordination plane (syn isomer) and the other with one H8 atom on each side of the coordination plane (anti isomer). Refine and save both using the MOMEC97 force field. Section 17.14 describes how to enforce planarity of a coordination compound. Two dimensional NMR methods can be used to determine which isomer dominates - as long as interconversion of the isomers is not rapid on the NMR time scale. The data used here are hypothetical and we have assumed that one isomer dominates to the exclusion of the other and that there is no interconversion, i. e., the observed NMR spectrum is that of an isomerically pure compound. 

Figure 20a is a plot of the data on semilogarithmic axes the potential is on a linear scale, whereas the current is on a logarithmic scale. The data form a straight line with a slope of b, referred to as the Tafel slope, in honor of Henrik Tafel, who studied the HER in the early 1900s (20). The data shown have a Tafel... 

Scale the data along the rows, by making each row add up to 1. Perform PCA. Why is the resultant scores plot of little physical meaning ... 

Scale the data in this region, so that each row is of a constant total. 

In Figure 7-2, four sets of shear rate versus shear stress data are shown on a double logarithmic scale. The data of the soup intersected and crossed over the data of... 

Figure 5 shows the plot of RMS radius versus MW on a log-log scale. The data for the elution volume lower than 3 mL were dropped... 

As shown in Figure 17 for the IR band near 563 cm, there is a strong change of dat/dTat To 825 K which is associated with a structural phase transition. Scaling the data at 7 > 500 K, the temperature dependencies of 5o(7) andQ (T) can be written as... 

If the ANNs present numerous advantages, they also present some weaknesses. Indeed they are versatile tools, but they require some experience to be correctly used to correctly scale the data, tune the network, and avoid overtraining and overfitting. More complete critical reviews of the ANN models can be found elsewhere [2,97-100],... 

If the variables have been measured in different units, then it may be necessary to scale the data to make the values comparable. " An equivalent procedure is to compute a weighted Euclidean distance. 

Fig. 5.9 Molal chemical shifts for the alkali metal and alkaline earth metal chlorides in aqueous solution plotted against the polarizing strength of the cation estimated as zjr (Pauling scale). The data shown as (o) have been corrected for the contribution of CP, and represent estimates of the cation contribution (see text).

---