Results/statistical packages analysis

D. Analytical Testing Methods PROTOCOL EXECUTION ANALYSIS OF RESULTS/STATISTICAL PACKAGES DOCUMENTATION OF RESULTS ANALYST CERTIFICATION AND TRAINING TRANSFEROR TECHNICAL OWNERSHIP... 

Cluster analysis is far from an automatic technique each stage of the process requires many decisions and therefore close supervision by the analyst. It is imperative that the procedure be as interactive as possible. Therefore, for this study, a menu-driven interactive statistical package was written for PDP-11 and VAX (VMS and UNIX) series computers, which includes adequate computer graphics capabilities. The graphical output includes a variety of histograms and scatter plots based on the raw data or on the results of principal-components analysis or canonical-variates analysis (14). Hierarchical cluster trees are also available. All of the methods mentioned in this study were included as an integral part of the package. 

Quantitative methodology uses large or relatively large samples of subjects (as a rule students) and tests or questionnaires to which the subjects answer. Results are treated by statistical analysis, by means of a variety of parametric methods (when we have continuous data at the interval or at the ratio scale) or nonparametric methods (when we have categorical data at the nominal or at the ordinal scale) (30). Data are usually treated by standard commercial statistical packages. Tests and questionnaires have to satisfy the criteria for content and construct validity (this is analogous to lack of systematic errors in measurement), and for reliability (this controls for random errors) (31). 

Several statistics for multivariate tests are known from the literature [AHRENS and LAUTER, 1981 FAHRMEIR and HAMERLE, 1984] the user of statistical packages may find several of them implemented and will rely on their performing correctly. Other, different, tests for separation of groups are used to determine the most discriminating results in discriminant analysis with feature reduction. 

The G-BASE project has used several statistical packages to perform this nested ANOVA analysis (e.g., Minitab and SAS). It currently uses an MS Excel procedure with a macro based on the equations described by Sinclair (1983) in which the ANOVA is performed on results converted to logio (Johnson, 2002). Ramsey et al. (1992) suggest that the combined analytical and sampling variance should not exceed 20% of the total variance with the analytical variance ideally being <4%. 

Standard statistical packages for computing models by least-squares regression typically perform an analysis of variance (ANOVA) based upon the relationship shown in Equation 5.15 and report these results in a table. An example of a table is shown in Table 5.3 for the water model computed by least squares at 1932 nm. 

With most statistics packages, data that are to be subjected to a one-way analysis of variance are entered into two columns in a similar way to that seen with a two-sample f-test (Section 6.8). One column contains a series of codes indicating what catalyst was used and the other column contains the corresponding experimental results. In the first five rows, the results are labelled as being due to the use of platinum (Pt), the next five are due to palladium (Pd) and so on. The general appearance will be as in Table 13.2. 

In most statistical packages, the implementation of the analysis of variance includes an option to select a Tukey s test. The format of the output varies enormously, but (as in Table 13.4) should include a list of confidence intervals for the difference between each possible pair of catalysts. Each line of output shows the difference calculated as the yield with the first metal minus that with the second. The results are shown ordered according to yield [palladium (highest) to platinum (lowest)]. 

The book is aimed at those who have to use statistics, but have no ambition to become statisticians per se. It avoids getting bogged down in calculation methods and focuses instead on crucial issues that surround data generation and analysis (sample size estimation, interpretation of statistical results, the hazards of multiple testing, potential abuses, etc.). In this day of statistical packages, it is the latter that cause the real problems, not the number-crunching. 

A wide range of preprogrammed search facilities is available, with all sorts of selection modes (by report number, time period, unit, status, causal code, and any combination of these). From these analyses, output files may be generated for more advanced statistical packages. Graphics facilities enable analysis results to be displayed as pie-charts, etc., for easy and fast interpretation. 

Pooling of Estimates Following the approach used in the MI paradigm, after M supplementations have been created for a data set, they are then analyzed using a standard PK/PD or statistical package. There are now M completed data sets containing the observed values and the supplemented values instead of one. The PK/PD or statistical analysis must be done M times, once on each complete data set. Across M data sets the results will vary, reflecting the uncertainty due to supplemental observations. The M complete data analyses are combined to create one repeated-supplementation inference. 

It is not possible to perform three-way or higher ANOYA in Excel, nor is it practical to perform the calculations manually. Many statistical packages do offer such analysis and require the data to be in a somewhat different form. The measurement results are in one column (sometimes known as the dependent variable) and each factor is represented by another column in which the level is given. Up to now we have considered situations in which the different levels of a factor are discrete entities—analysts, methods, etc. However, we have also referred to factors that are continuous variables, such as time and temperature. The model that ANOYA builds in each case is slightly different, and most software can cope with this. The output from different software programs varies but mostly contains the important information of the mean squares, F values and associated probabilities. 

Five parameters in the data-set were found to be unchanged for all 35 compounds and removed from the matrix. These parameters are H-DO for positions II, IV and V and H-AC for positions IV and V. After the redundant elements had been removed, the resulting [35x47] matrix was correlated to the vector of the biological activity. To perform the linear stepwise regression analysis, the STEPWISE procedure of the SAS statistical package ( ) and BASIC programs were used. 

Data for MM compositions of different SWMs and their leachates were examined statistically in order to determine any significant compositional variations among samples. Most statistical analyses were performed using the SAS Statistical Package V 6.12 [335]. In this report, the results of Q-mode factor analysis and Unear programming techniques will be presented. The objectives of the statistical analyses were to define the MM characteristics for the different SWMs and their leachates, and to determine their original sources. 

It should be noted that, owing to the nature of molecular Leld analysis, there would be hundreds or thousands of descriptors forthe molecular steric and electrostatic force Lelds (Table 3.5), and it is inconvenient and unnecessary to write down the QSSR equation in a CoMFA study since these descriptors are generated and directly used to predict targeted properties using the same software package. Statistical results forthe Lnal models correlating the mole fraction solubility in isopropyl alcohol with the desired parameters for 60 aromatic and heteroaromatic crystalline compounds are... 

---