Randomization classifying data

Count data, based on a random selection of individuals or items which are classified according to two different criteria, can be statistically analyzed through the distribution. The purpose of this analysis is to determine whether the respective criteria are dependent. That is, is the product preferred because of a particular characteristic ... 

In order to classify promiscuous and selective compounds, we used the NB modeling protocol available in Pipeline Pilot (Scitegic) [53]. The data was split randomly into 5193 compounds for modeling and 574 compounds for testing the models. In addition to the test set, 302 known drugs were also profiled and kept separate for testing the models. All sets were checked visually to ensure that no chemical classes were overrepresented in one set or the other. 

Two methods are used to evaluate the predictive ability for LDA and for all other classification techniques. One method consists of dividing the objects of the whole data set into two subsets, the training and the prediction or evaluation set. The objects of the training set are used to obtain the covariance matrix and the discriminant scores. Then, the objects of the training set are classified, so obtaining the apparent error rate and the classification ability, and the objects of the evaluation set are classified to obtain the actual error rate and the predictive ability. The subdivision into the training and prediction sets can be randomly repeated many times, and with different percentages of the objects in the two sets, to obtain a better estimate of the predictive ability. 

The remaining errors in the data are usually described as random, their properties ultimately attributable to the nature of our physical world. Random errors do not lend themselves easily to quantitative correction. However, certain aspects of random error exhibit a consistency of behavior in repeated trials under the same experimental conditions, which allows more probable values of the data elements to be obtained by averaging processes. The behavior of random phenomena is common to all experimental data and has given rise to the well-known branch of mathematical analysis known as statistics. Statistical quantities, unfortunately, cannot be assigned definite values. They can only be discussed in terms of probabilities. Because (random) uncertainties exist in all experimentally measured quantities, a restoration with all the possible constraints applied cannot yield an exact solution. The best that may be obtained in practice is the solution that is most probable. Actually, whether an error is classified as systematic or random depends on the extent of our knowledge of the data and the influences on them. All unaccounted errors are generally classified as part of the random component. Further knowledge determines many errors to be systematic that were previously classified as random. 

Example 1 A medical device manufacturer is concerned about the nonconforming (defective) and the nonconformity (defect) produced in its recently set-up production line. Twenty batches of this medical device were randomly selected from the production line. Each batch contained 100 units. Each unit is inspected and is classified as either conforming or nonconforming. During the inspection, the number of nonconformities (defects) was also counted. The data collected are shown in Table 3. 

Data generally are classified as either deterministic or random. Deterministic means that the process under study can be described by an explicit mathematical relationship. Random means that the phenomenon under study cannot be described by an explicit mathematical function because each observation of the phenomenon is unique. A single representation of a random phenomenon is called a sample function. If, as in all experiments, the sample function is of finite length, then it is called a sample record. The set of all possible sample functions, y r), which the random process might produce, defines the random or stochastic process. The mean value and autocorrelation function for a random process are defined by... 

Uncertainty is classified in two major groups Random error is always present in experimental data and can never be completely eliminated. It can result from the random nature of collisions that lead to chemical or biochemical reactions, or may be caused by small voltage fluctuations in measurement instrumentation. Random error causes positive and negative deviations from the true value, and affects the precision of the results. Precision is usually discussed in terms of standard deviation (,v) and relative standard deviation (RSD), both defined later in this chapter. Systematic error is produced by a more or less constant mistake, and results in a... 

Conventional CPUs found in the majority of modern computers, such as those manufactured by Intel and advanced micro devices (AMD), are designed for sequential code execution as per the Von Neumann architecture [16]. While running a program, the CPU fetches instructions and associated data from the computer s random access memory (RAM), decodes it, executes it, and then writes the result back to the RAM. Within the realm of Flynn s taxonomy [17], this would be classified as single instruction, single data (SISD). 

Gao et al. [36] used binary QSAR based on topological descriptors and indicator variables (including one for the phenolic hydroxyl group) to derive a classification model that separates active from inactive compounds. The model was trained on 410 diverse molecules, and it demonstrated its predictive power on a test set of 53 randomly selected molecules from which 94% were correctly classified. The biological data were selected from four different laboratories, so there might be some inconsistency with respect to the classification of the model. 

---