摘要:

The use of hidden Markov models for speech recognition has become predominant in the last several years, as evidenced by the number of published papers and talks at major speech conferences. The reasons this method has become so popular are the inherent statistical (mathematically precise) framework; the ease and availability of training algorithms for cstimating the parameters of the models from finite training sets of speech data; the flexibility of the resulting recognition system in which one can easily change the size, type, or architecture of the models to suit particular words, sounds, and so forth; and the ease of implementation of the overall recognition system. In this expository article, we address the role of statistical methods in this powerful technology as applied to speech recognition and discuss a range of theoretical and practical issues that are as yet unsolved in terms of their importance and their effect on performance for different system implementations.

ISSN:0040-1706    

DOI:10.1080/00401706.1991.10484833

出版商:Taylor & Francis Group    

年代:1991

数据来源: Taylor

摘要:

This article discusses methods whereby reports of warranty claims can be used to estimate the expected number of warranty claims per unit in service as a function of the time in service. These methods provide estimates that are adjusted for delays or lags corresponding to the time from the claim until it is entered into the data base used for analysis. Forecasts of the number and cost of claims on the population of all units in service are also developed, along with standard errors for these forecasts. The methods are based on a log-linear Poisson model for numbers of warranty claims. Both the case of a known distribution of reporting lag and simultaneous estimation of that distribution are considered. The use of residuals for model checking, extensions to allow for extra-Poisson variation, and the estimation of warranty costs are also considered.

ISSN:0040-1706    

DOI:10.1080/00401706.1991.10484834

出版商:Taylor & Francis Group    

年代:1991

数据来源: Taylor

摘要:

Component-test plans are often designed by allocating system reliability among the system's components, then choosing individual component plans suitable for demonstrating achievcment of each component's reliability goal. This approach does not consider how much information relative to the system reliability goal is provided by the ensemble of component tests. We consider the notion of system reliability operating characteristic (OC) curves, based on the component tests, and illustrate their use in designing or evaluating an overall test program. By specifying OC values (akin to producer's and consumer's risks), optimum, system-ortented component-test plans can be derived. These ideas are illustrated for a series system, and for a simple series-parallel system, with binomial data.

ISSN:0040-1706    

DOI:10.1080/00401706.1991.10484835

出版商:Taylor & Francis Group    

年代:1991

数据来源: Taylor

摘要:

Exponentially weighted moving average (EWMA) quality-monitoring schemes capahlc of detecting changes in both location and spread, referred to as omnibus EWMA's. are proposed. Omnibus EWMA's are based on the exponentiation of the absolute value of the standardized sample mean of the observations. The process target mean and standard deviation are used for standardizing. Design procedures and considerations, including a fast initial response feature, are discussed. Average run lengths for exponents of .5 and 2 and various values of the weighting constant are presented. The proposed schemes are compared with each other and with a number of selected, well-known omnibus schemes and a proposed omnibus cumulative sum scheme. Schemes with different sample sizes are used, and off-target models include both static and dynamic alternatives. Comparisons are based on average number of observations to detection and expected loss due to poor quality. One of the omnibus EWMA schemes is shown to be best with regard to expected loss if a uniform prior is placed on the off-target models studied.

ISSN:0040-1706    

DOI:10.1080/00401706.1991.10484836

出版商:Taylor & Francis Group    

年代:1991

数据来源: Taylor

摘要:

One of the fundamental purposes of response surface methodology is to develop a relatively simple linear model to explain a more complex phenomenon. Since the proposed model is only an approximation, analysts almost always must address the potential of bias due to model misspecification. This article proposes a graphical approach for evaluating response surface designs in terms of the mean squared error of prediction. In particular, this article develops measures of the maximum and minimum mean squared error of prediction over the surface of spheres through a region of interest. These measures are then plotted against radius. The resulting plots reveal at a glance how a response surface design performs in the presence of model misspecification. These plots are used to investigate the performances of several common response surface designs.

ISSN:0040-1706    

DOI:10.1080/00401706.1991.10484837

出版商:Taylor & Francis Group    

年代:1991

数据来源: Taylor

摘要:

For the case in which two independent samples arc to be compared using a nonparametric test for location shift, we propose a bootstrap technique for estimating the sample sizes required to achieve a specified power. The estimator (called BOOT) uses information from a small pilot experiment. For the special case of the Wilcoxon test, a simulation study is conducted to compare BOOT to two other sample-size estimators. One method (called ANPV) is based on the assumption that the underlying distribution is normal with a variance estimated from the pilot data. The other method (called NOETHER) adapts the sample size formula of Noether for use with a location-shift alternative. The BOOT and NOETHER sample-size estimators are particularly appropriate for this nonparametric setting because they do not require assumptions about the shape of the underlying continuous probability distribution. The simulation study shows that (a) sample size estimates can have large uncertainty, (b) BOOT is at least as accurate as and can be much more accurate than ANPV, and (c) BOOT and NOETHER achieve similar accuracy, although NOETHER is prone to underestimation.

ISSN:0040-1706    

DOI:10.1080/00401706.1991.10484838

出版商:Taylor & Francis Group    

年代:1991

数据来源: Taylor

摘要:

This article is concerned with procedures for detecting multipleyouthers in linear regression. A generalized extreme studentized residual (GESR) procedure, which controls type I error rate, is developed. An approximate formula to calculate the percentiles is given for large samples and more accurate percentiles forn≤ 25 are tabulated. The performance of this procedure is compared with others by Monte Carlo techniques and found to be superior. The procedure. however, fails in detectingyoutliers that are on high-leverage cases. For this. a two-phase procedure is suggested. In phase 1, a set of suspect observations is identified by GESR and one of the diagnostics applied sequentially. In phase 2, a backward testing is conducted using the GESR procedure to see which of the suspect cases are outlicrs. Several examples are analyzed.

ISSN:0040-1706    

DOI:10.1080/00401706.1991.10484839

出版商:Taylor & Francis Group    

年代:1991

数据来源: Taylor

摘要:

The idle column method, proposed by Taguchi, enables a two-level orthogonal array to be used as the basis of a mixed-level fractional factorial experiment. The resulting designs are not orthogonal but can in some circumstances be described as nearly orthogonal. The method must be applied with care, because an apparently minor change in the technique produces a different kind of design. We illustrate these points with a simple example. The same example is used to compare the “orthogonal” analysis of nonorthogonal designs, as advocated by Taguchi, with the least squares method. Finally, we examine some alternative designs and try to identify the circumstances in which the idle-column method is likely to be most useful.

ISSN:0040-1706    

DOI:10.1080/00401706.1991.10484840

出版商:Taylor & Francis Group    

年代:1991

数据来源: Taylor

ISSN:0040-1706    

DOI:10.1080/00401706.1991.10484841

出版商:Taylor & Francis Group    

年代:1991

数据来源: Taylor

ISSN:0040-1706    

DOI:10.1080/00401706.1991.10484842

出版商:Taylor & Francis Group    

年代:1991

数据来源: Taylor