Principal Component Approach Research Articles

Temporal analysis of human development indicators: principal component approach.

In this study, we used the weighted principal component method on the human development indicators to measure and analyzethe progress of human development in the world. The main principal component was employed to quantify the temporal changes of the human development of several selected countries by the proposed Z-test. The trends of the human development in the period of market transition in two large countries, China and Russia, were discussed in terms of the impact of public health as well as economic development. The association of the main principal component obtained from our study and the human development index reported by the United Nations Development Programme was estimated by the Spearman’s rank correlation coefficient.

D-ISODATA: A Distributed Algorithm for Unsupervised Classification of Remotely Sensed Data on Network of Workstations

With the advent of high-speed networks and the availability of powerful high-performance workstations, network of workstations has emerged as the most cost-effective platform for computation-intensive applications. One of the major applications for the network of workstations is in the field of remote sensing, where because of the high dimensionality of data, most of the existing data exploitation procedures are computation-intensive. To test the utility of the network of workstations in the field of remote sensing we have adopted a modified version of the well-known ISODATA classification procedure which may be considered as the benchmark for all unsupervised classification algorithms. The ISODATA algorithm is an iterative method that uses Euclidean distance as the similarity measure to cluster data elements into different classes. We have designed and developed a distributed version of ISODATA algorithm (D-ISODATA) on the network of workstations under a message-passing interface environment and have obtained promising speedup. To reduce the processing load and thereby increase the throughput, the ISODATA procedure is commonly applied to only the first few principal component images derived from the original set of the multispectral images. The drawback with the principal component approach is that it is based entirely on the statistical significance of the spectra, rather than the uniqueness of the individual spectra. As, small objects and ground features would likely manifest themselves in the last principal component images, that is, eigen images, discarding them prior to classification would lead to the loss of valuable information. The significant enhancement in processing speed on the network of workstations makes it possible for us to apply our distributed algorithm D-ISODATA to the entire set of multispectral images directly, thereby preserving all the spectral signatures in the data, regardless of their statistical significance.

Forecasting with unequally spaced data by a functional principal component approach

The Principal Component Regression model of multiple responses is extended to forccast a continuous-time stochastic process. Orthogonal projection on a subspace of trigonometric functions is applied in order to estimate the principal components using discrete-time observations from a sample of regular curves. The forecasts provided by this approach are compared with classical principal component regression on simulated data.

Eigenfeatures as intermediate-level representations: The case for PCA models

Eigenfeatures are created by the principal component approach (PCA) used on objects described by a low-level code (i.e., pixels, Gabor jets). We suggest that eigenfeatures act like the flexible features described by Schyns et al. They are particularly suited for face processing and give rise to class-specific effects such as the other-race effect. The PCA approach can be modified to accommodate top-down constraints.

Automatic quantification of immunohistochemically stained cell nuclei using unsupervised image analysis.

A method for quantification of images of immunohistochemically stained cell nuclei by computing area proportions is presented. The image is transformed by a principal component transform. The resulting first component image is used to segment the objects from the background using dynamic thresholding of the P2/A‐histogram, where P2/A is a global roundness measure. Then the image is transformed into principal component hue, defined as the angle around the first principal component. This image is used to segment positive and negative objects. The method is fully automatic and the principal component approach makes it robust with respect to illumination and focus settings. An independent test set consisting of images grabbed with different focus and illumination for each field of view was used to test the method, and the proposed method showed less variation than the intraoperator variation using supervised Maximum Likelihood classification.

Locating some types of random errors in Digital Terrain Models

The increasing use of Geographical Information System applications has generated a strong interest in the assessment of data quality. As an example of quantitative raster data, we analysed errors in Digital Terrain Models (DTM). Errors might be classified as systematic (strongly dependent on the production methodology) and random. The present work attempts to locate some types of randomly distributed, weakly spatially correlated errors by applying a new methodology based on Principal Components Analysis. The Principal Components approach presented is very different from the typical scheme used in image processing. A prototype implementation has been conducted using MATLAB, and the overall procedure has been numerically tested using a Monte Carlo approach. A DTM of Stockholm, with integer-valued heights varying from 0 to 59 m has been used as a testbed.The model was contaminated by adding randomly located errors, distributed uniformly within 4 m and 4m. The procedure has been applied using both spike shaped (isolated errors) and pyramid-like errors. The preliminary results show that for the former, roughly half of the errors have been located with a Type I error probability of 4.6 per cent on average, checking up to 1 per cent of the dataset. The associated Type II error of the larger errors (of exactly 4m or 4 m) drops from an initial value of 1.21 per cent down to 0.63 per cent. By checking another 1 per cent of the dataset, such error drops to 0.34 per cent implying that about 71 per cent of the 4m errors have been located; Type I error was below 11.27 per cent. The results for pyramid-like errors are slightly worse, with a Type I error of 25.80 per cent on average for the first 1 per cent effort, and a Type II error drop from an initial value of 0.81 per cent down to 0.65 per cent. The procedure can be applied both for error detection during the DTM generation and by end users. It might also be used for other types of quantitative raster data.

The choice of principal variables for computing the Physical Quality of Life Index

213 The present paper addresses the issue of choice of principal variables for computing the Physical Quality of Life Index (PQLI). To this end, the variable selection strategy considered by Jolliffe, which utilizes the principal components approach, is applied to the data for 150 countries reported by Morris. To test the robustness of our results the methodology is also applied to four different subsamples of the 150 countries (low income, lower middle income, upper middle income and high income). All the results indicate the need to retain only one of the three components of the PQLI, namely the literacy index. Thus, a simpler PQLI series based solely on the literacy index would be cheaper to compute without loss of too much information.

A principal component approach to dynamic regression models

In this paper we introduce a dynamic regression model that states how an output is related to an input allowing future values forecasting. The basic tools to set up this model are the orthogonal decomposition of a discrete time stochastic process by means of its principal components analysis, and the linear regression performed on the principal components of input and output processes. The behaviour of this model is empirically studied on real data, showing that low forecast errors are obtained by using this model. A comparison between such a model and a transfer function one, for a particular two time-series case, is discussed.

Multipoint dissolution specification and acceptance sampling rule based on profile modeling and principal component analysis

In dissolution testing multiple dissolution measurements at specific time points are needed in quality control when the compliance of the product requires controlled dissolution throughout the time course. The dissolution specification based on general multivariate confidence region was proposed by Chen and Tsong (8). This paper presents two alternative procedures when the dissolution profile consists of important measurements at more than 4 time points. In the first procedure, when the dissolution profile can be described by a physical curve through modeling, the dissolution specification is developed based on the confidence region of the parameters of the physical curve. In the second procedure, the principal components (PCS) as the linear combinations of the dissolution measurements are identified and dissolution specification is set based by the confidence intervals of the values of principal components. In both approaches the specification can be set at lower dimensions than the general multivariate confidence region approach. A single-stage acceptance rule can be used in both approaches by first projecting the dissolution values of each tablet in the new testing batch onto the determined parameters axes (through modeling in modeling approach and through projection on the selected PCS in principal component approach). Then check if the projections of the new tablet fall within the specifications. Finally, count the number of tablets that fall outside the specification limits and reject the batch if the proportion of out-of-specification tablet is high and accept the lot for release if the proportion is low.

AN OPTIMALITY CRITERION FOR AGGREGATING A SET OF TIME SERIES IN A COMPOSITE INDEX

Linear combination is a common feature of most methods designed to aggregate a set of time series into a composite index. In this paper the principal components approach is utilized by means of spectral techniques to improve the variance explained by the first principal component. The optimality criterion resorts to the alignment of the elements of the vector time series.

INDIVIDUAL VARIATION OF ONTOGENIES: A LONGITUDINAL STUDY OF GROWTH AND TIMING.

This study of growth and developmental time in the water strider Limnoporus canaliculatus (Heteroptera: Gerridae) is based on longitudinal data from specimens reared individually in the laboratory. I analyzed multivariate allometry using a common principal components approach. This technique identified patterns of variation that were uncorrelated both within and among instars and which remained fairly constant throughout the growth period; in contrast, the overall amount of variation increased from young to older instars. Negative correlations between size and subsequent growth increments indicated convergent growth in the first three instars, but there was a transition to positive correlations (divergent growth) in later instars. Analysis of covariation among measurements made in different instars showed strong ontogenetic autocorrelation and revealed patterns remarkably similar to those found in mammals and birds; yet corresponding analyses of growth increments showed mainly independent variation in different instars. Therefore, I conclude that the strong correlations among stage-specific measurements result from the part-whole relationships inherent to these cumulative size data, but do not reflect specific properties of the organisms studied. In contrast to size increments, instar durations of water striders were highly correlated throughout the larval period, indicating that individuals tended to develop at either relatively fast or relatively slow rates in all instars. The correlations between growth increments and instar durations were nil or negative, contrary to expectations from life-history theory. The results of these analyses of individual variation match the findings from other water striders and from interspecific comparisons in the genus Limnoporus, but information about physiological mechanisms of molting and growth in insects cannot completely explain the patterns observed.

Statistical process monitoring with principal components

Most industrial processes are characterized by a system of several variables, all of which are subject to drifts, disturbances, and assignable causes of variation. In the chemical and process industries, there are often inertial forces arising from raw material streams, reactors and tanks that introduce serial correlation over time into these variables. This autocorrelation can have a profound impact on the effectiveness of the statistical monitoring methods used for such processes. This paper reviews some of the available methodology for multivariate process monitoring and shows the effectiveness of principal components in this context. An application of the principal components approach with correlated observation vectors is presented. The effectiveness of this procedure to indicate process upsets is discussed.

STATISTICAL PROCESS MONITORING WITH PRINCIPAL COMPONENTS

Most industrial processes are characterized by a system of several variables, all of which are subject to drifts, disturbances, and assignable causes of variation. In the chemical and process industries, there are often inertial forces arising from raw material streams, reactors and tanks that introduce serial correlation over time into these variables. This autocorrelation can have a profound impact on the effectiveness of the statistical monitoring methods used for such processes. This paper reviews some of the available methodology for multivariate process monitoring and shows the effectiveness of principal components in this context. An application of the principal components approach with correlated observation vectors is presented. The effectiveness of this procedure to indicate process upsets is discussed.

A comparison of methods for determining dimensionality in Rasch measurement

This study compares the Rasch item fit approach for detecting multidimensionality in response data with principal component analysis without rotation using simulated data. The data in this study were simulated to represent varying degrees of multidimensionality and varying proportions of items representing each dimension. Because the requirement of unidimensionality is necessary to preserve the desirable measurement properties of Rasch models, useful ways of testing this requirement must be developed. The results of the analyses indicate that both the principal component approach and the Rasch item fit approach work in a variety of multidimensional data structures. However, each technique is unable to detect multidimensionality in certain combinations of the level of correlation between the two variables and the proportion of items loading on the two factors. In cases where the intention is to create a unidimensional structure, one would expect few items to load on the second factor and the correlation be...

The Intensive Care Unit Medical Director as Manager Impact on Performance

This article examines the impact of greater management involvement by the medical director on efficiency of bed allocation in the intensive care unit (ICU) or critical care unit. Managerial involvement is modeled using a principal components approach in terms of perceived supervision, conflict resolution regarding bed allocation at critical times, extent of control over treatment, and employment status. Using data from a 1991 survey of 2,879 ICUs in 1,706 hospitals conducted by the Society of Critical Care Medicine, two equations reflecting efficiency--the ICU occupancy rate and the presence of misallocated ICU patients--were estimated. It was found that greater involvement by medical directors in the day-to-day management of the ICU significantly reduces the average occupancy rate in ICUs and also the probability of patients misallocated to the ICU, suggesting superior resource allocation in ICUs as a result. These results also suggest that the managerial impact of the medical director is greater in ICUs in high-occupancy hospitals.

More about the difference between men and women: evidence from linear neural networks and the principal-component approach.

The ability of a statistical/neural network to classify faces by sex by means of a pixel-based representation has not been fully investigated. Simulations with pixel-based codes have provided sex-classification results that are less impressive than those reported for measurement-based codes. In no case, however, have the reported pixel-based simulations been optimized for the task of classifying faces by sex. A series of simulations is described in which four network models were applied to the same pixel-based face code. These simulations involved either a radial basis function network or a perceptron as a classifier, preceded or not by a preprocessing step of eigendecomposition. It is shown that performance comparable to that of the measurement-based models can be achieved with pixel-based input (90%) when the data are preprocessed. The effect of the eigendecomposition preprocessing of the faces is then compared with spatial-frequency analysis of face images and analyzed in terms of the perceptual information it captures. It is shown that such an examination may offer insight into the facial aspects important to the sex-classification process. Finally, the contribution of hair information to the performance of the model is evaluated. It is shown that, although the hair contributes to the sex-classification process, it is not the only important contributor.

Synthetic images by subspace transforms. I. Principal components images and related filters.

The principal component (PC) approach offers compressions of an image sequence into fewer images and noise suppressing filters. Multiple MR images of the same tomographic slice obtained with different acquisition parameters (i.e., with different TR, TE, and flip angles), time sequences of images in nuclear medicine, and cardiac ultrasound image sequences are examples of such input image sets. In this paper noise relationships of original and linearly transformed image sequences in general, and specifically of original, PC, and PC-filtered images are discussed. As the spinoff, it introduces locally weighted PC transforms and filters, nonlinear PC's, and a single-image based filter for suppression of noise. Examples illustrate increased perceptibility of anatomical/functional structures in PC images and PC-filtered images, including extraction of physiological functional information by PC loading curves. Generally, the more correlated the original images are, the more effective is the PC approach.

Application of the scaled subprofile model to functional imaging in neuropsychiatric disorders: A principal component approach to modeling brain function in disease

AbstractRecent advances in functional neuroimaging have presented a challenge to traditional statistical methods in characterizing the effects of neuropsychiatric illness on brain function. The most common approach for analyzing regional group differences has relied on t‐tests with significance thresholds selected to reduce the potential effect of multiple statistical tests. Regional covariance analysis offers an alternative to this threshold‐based, group difference approach by identifying the functional interactions among brain regions that can be spatially distributed throughout the brain. The Scaled Subprofile Model (SSM) is one form of regional covariance analysis that has been applied to the study of patient groups. Based on a modified principal component analysis, the SSM offers a method for modeling regionally specific patterns of brain function whose expression can be evaluated between groups and validated against clinical measures of patient disease severity and neuropsychological test scores. We review the application of the SSM, to date, in studies of the effects of neurological and psychiatric illness on brain function, including a discussion of SSM methodology and its application to the study of resting state functional neuroimaging in patient groups. SSM analyses applied to studies of Alzheimer's disease, Parkinson's disease, major depressive disorder, AIDS dementia complex, and neoplastic disease each identified functionally specific topographic effects that were associated with clinical disease severity. The results of the SSM analyses suggest that neuropsychiatric disorders may alter functional networks or systems of neural activity in ways that can be expressed as regional covariance patterns in resting functional imaging data. © 1994 Wiley‐Liss, Inc.

Evaluation of alternative leading indicators of British Columbia industrial employment

Leading indicators based on principal components and weighted r 2 values are evaluated in this study. The tests are based on out-of-sample forecasts of British Columbia employment. Overall, the principal component approach is found to be superior in tests conducted separately by forecast horizons ranging from 1 to 12 months, and in tests conducted in the months immediately following the 1984–1992 turning point. This finding is explained in the study by the methodological superiority of the leading indicator based on principal components of the input series.

Minimum sensitivity IIr filter design using principal component approach

Minimum sensitivity IIr filter design using principal component approach