摘要:

The application of optimal experimental design theory to models for dynamic systems is surveyed. Preliminary sections briefly discuss the models used and the main points of statistical optimal design theory. Then the ways in which the latter carry over to dynamic models are described. These applications are split roughly into those involving choice of input functions and those in which sampling rates also are selected.

ISSN:0040-1706    

DOI:10.1080/00401706.1980.10486160

出版商:Taylor & Francis Group    

年代:1980

数据来源: Taylor

摘要:

We develop a family of computer search methods for finding optimum designs, that generalize and improve upon Mitchell's exchange and DETMAK algorithms. Our emphasis is on time- and space-saving considerations that permit us to handle larger problems, or to run each search “try” for the optimum more rapidly, than was previously possible. This last means that more tries can be attempted for a given total cost, with consequent greater chance of finding an optimum or near-optimum design. Indeed, we have found a number of new optimum or improved designs using these methods. Fork= 6 to 12 parameters and withnobservations andk≤n≤ 2k, our methods are typically 15 to 50 times faster than DETMAX (more asnandkincrease), with comparable success rates. Numerical studies in linear and quadratic regression examples treat also the effect of amount of initial randomization on the success of a try.

ISSN:0040-1706    

DOI:10.1080/00401706.1980.10486161

出版商:Taylor & Francis Group    

年代:1980

数据来源: Taylor

摘要:

An empirical comparison of existing algorithms for the computer generation of exact D-optimal experimental designs is carried out. Among algorithms considered were those due to Wyrnn, Mitchell, Fedorov, and Van Schalkwyk. A procedure for rounding off approximate designs as suggested by Kiefer is also evaluated. A modification of the Fedorov algorithm is given and shown to effect substantial decreases in the computer time required for design generation.

ISSN:0040-1706    

DOI:10.1080/00401706.1980.10486162

出版商:Taylor & Francis Group    

年代:1980

数据来源: Taylor

摘要:

Maximum likelihood (ML) methods are described for the identification of outliers in single sample or regression problems, based on mixture models. The EM algorithm provides a simple and easily programmed iterative solution for the ML estimates of the parameters in the models. The procedure is illustrated on three examples.

ISSN:0040-1706    

DOI:10.1080/00401706.1980.10486163

出版商:Taylor & Francis Group    

年代:1980

数据来源: Taylor

摘要:

A method is presented based on augmented data sets for combining biased and robust regression techniques. The estimates are constrained robust estimates, using an appropriately chosen ridge, shrinkage or principal components constraint. Examples are provided to illustrate the ability of the procedure to shrink the estimated coefficients and to automatically detect and discount the effects of observations with large random error components.

ISSN:0040-1706    

DOI:10.1080/00401706.1980.10486164

出版商:Taylor & Francis Group    

年代:1980

数据来源: Taylor

摘要:

In ap-variate weighted least squares multiple regression based onnpoints, each of thepregression coefficients can be calculated from a univariate weighted regression through the origin usingN= (np) composite points. This result can be applied to (1) display thep-dimensional multiple regression aspunivariate regressions each based onNcomposite points, (2) isolate collections of points with large leverage, and (3) characterize the region of regression coefficients that can be obtained from fixed data with varying weights. Consequently, composite points are diagnostically useful in the context of robust regression.

ISSN:0040-1706    

DOI:10.1080/00401706.1980.10486165

出版商:Taylor & Francis Group    

年代:1980

数据来源: Taylor

摘要:

Whenstepwise dkriminantanalysis or MANOVA is performed, the mean and percentage points of Wii A are shifted downward. This bias was examined by Monte Carlo methods for the case of no differences between groups. The downward bias is especially pronounced when the number of variables exceeds the degrees of freedom for error.

ISSN:0040-1706    

DOI:10.1080/00401706.1980.10486166

出版商:Taylor & Francis Group    

年代:1980

数据来源: Taylor

摘要:

Methods are given for constructing, from past sample data, prediction intervals (i) to contain the ratio of the variances of future samples from two normal distributions and (ii) to contain the ratio of the means from future samples from two exponential distributions. Such intervals are required, for example, when a producer wishes to predict the results of a future test, to be conducted by a consumer or regulatory agency, to compare its product with a competitor's. A new tabulation of percentage points of the distribution of two independent F ratios to obtain such intervals is provided; this extends existing tabulations.

ISSN:0040-1706    

DOI:10.1080/00401706.1980.10486167

出版商:Taylor & Francis Group    

年代:1980

数据来源: Taylor

摘要:

In the replicated two-way random-effects model, the “Sometimes Pool” (SP) procedure is employed for the specification of the appropriate model and the subsequent test of the main effects. The “Sometimes Switch” (SW) procedure is introduced and shown to be a limiting form of SP as the cell size increases. This article concerns itself with the question of whether to pool interaction and error mean squares in order to get better tests for the main effects. The significance levels and the powers of the “sometimes” procedures are compared with those of the “Never Pool” (NP). The SP and SW procedures rarely have control over the main test significance level. We recommend modifying the experimental design, when possible, to maximize the interaction degrees of freedom and employing the “Never Pool” procedure.

ISSN:0040-1706    

DOI:10.1080/00401706.1980.10486168

出版商:Taylor & Francis Group    

年代:1980

数据来源: Taylor

摘要:

Consider the three-factor crossed classitkation components-of-variance model with interaction given by

ISSN:0040-1706    

DOI:10.1080/00401706.1980.10486169

出版商:Taylor & Francis Group    

年代:1980

数据来源: Taylor