Statistical methods mixture experiments

The excessive number of animals per test group is also pointed out and tackled with recommendations on the reduction of animal numbers. In particular, a better use of statistics in the design of experiments is recommended in order to reach an optimal compromise between animal number and variability of the results. The use of homogenous populations is advocated as a means to minimize interindividual variability if physiological variation between individual animals can be controlled, and statistical methods used to exploit this control to the full, the number of animals necessary for assay purposes can be dramatically reduced. This results, for instance, in the use of only one breed of rats for one set of tests, rather than a mixture of different breeds with the objective of mimicking the phenotypic variability of humans. This latter approach would result mainly on the study of the differences between breeds rather than the actual effects of the compound. 

A detailed analysis of the approaches and methods used in experiments with mixtures can be found in Cornell (1990). The multivariate statistical methods, such as multivariate hypothesis testing, principal component analysis, cluster analysis, and discriminant analysis are discussed in Dillon and Goldstein (1984). 

Traditional confirmatory statistical methods are not always the best option for analysis of chemical mixture experiments. A mixture experiment is defined as one where the response is dependent upon the relative proportion of ingredients comprising the mixture rather than the absolute amount of different ingredients that influence a response (/). Mixture analysis data can t be analyzed using conventional statistical techniques. Traditional methods assume independent results for each ingredient, an assumption which is violated by the use of relative mixtures. 

The statistical methods discussed thus far are of a quite general nature, routinely finding application beyond the bounds of the chemical industry. In this section, we will briefly highlight two statistical methodologies whose most important applications are to chemical problems. That is, we will touch on some of the ideas of mixture experiments and the role of statistics in mechanistic modeling. 

Final state analysis is where dynamical methods of evolving states meet the concepts of stationary states. By their definition, final states are relatively long lived. Therefore experiment often selects a single stationary state or a statistical mixture of stationary states. Since END evolution includes the possibility of electronic excitations, we analyze reaction products in terms of rovibronic states. 

Selected ion monitoring can be used for the determination of the relative amount of each component of a mixture, introduced into the mass spectrometer by the direct inlet probe However, such a determination requires reference mixtures of known composition for calibration. In the present experiment, since the monochloro pentaziridino derivative had not yet been isolated in the pure form, it was necessary to determine its concentration, by an auxiliary method, in a sample which could then be utilized as a reference mixture for further experiments. In order to do this we titrated chlorine in the toxic sample of MYKO 63 (B) by the classical method. The results indicated that the amount of N3P3AZJCI was between 0.5-1.5 %. The large statistical error is due to the low chlorine content in the sample examined. Thus, we used the remarkable possibilities provided by neutron activation analysis when the impurity to be quantified is a chlorinated moiety. It is well-known indeed that the C1 -f 2n peak is amongst the most easily detectable by neutron... 

Equations of state are used in engineering to predict the thermodynamic properties in particular the phase behaviour of pure substances and mixtures. However, since there is neither an exact statistical-mechanical solution relating the properties of dense fluids to their intermolecular potentials, nor detailed information available on intermolecular potential functions, all equations of state are, at least partially, empirical in nature. The equations of state in common use within both industry and academia are described elsewhere in this book and can be arbitrarily classified as follows (1), cubic equations derived from the observation of van der Waals that are described in Chapter 4 (2), those based on the virial equation discussed in Chapter 3 (3), equations based on general results obtained from statistical mechanics and computer simulations mentioned in Chapter 8 and (4), those obtained by selecting, based on statistical means, terms that best represent the available measurements obtained from a broad range of experiments as outlined in Chapter 12. The methods used for mixtures are also alluded to in these chapters and in Chapter 6. 

Conventional NMR deals with a large ensemble of spins. It means that the state of the system is in a statistical mixture, which is obviously inadequate for QIP. However, the NMR ability for manipulating spins states worked out by Cory et al. [24] and Chuang et al. [23] resulted in elegant methods for creating the so called effectively pure or pseudo-pure states. Behind the idea of the pseudo-pure states is the fact that NMR experiments are only sensitive to the traceless deviation density matrix. Thus, we might search for transformations that, applied to the thermal equilibrium density matrix, produce a deviation density matrix with the same form as a pure state density matrix. Once such state is created, all remaining unitary transformations will act only on such a deviation density matrix, which will transform as a true pure state. 

---