Statistical Measures of Parameter Estimates from Models Fit to Respiratory Impedance Data: Emphasis on Joint Vanabilities

To describe respiratory mechanical impedance data, many investigators have proposed electromechanical models and then fit them to data using formal parameter estimation techniques. This approach has resulted in confusion as to how to interpret the resulting estimated values, and hence as to which model is most appropriate. A key cause of this confusion is that most studies rely on the quality of fit between the model and the data as the only measure of model validity rather than performing adequate statistical analysis of the parameter estimates themselves. This paper describes several statistical measures that should be applied to parameter estimates obtained from forced oscillation data. Specifically, we describe standard errors of the parameter estimates, confidence intervals for each parameter estimate, and the joint confidence region for the parameters. Much emphasis is placed on the joint confidence region which, unlike the interval, allows for simultaneous variations in parameters. The measures are applied to an often used six-element model for respiratory impedance data of dogs from 4 to 64 Hz. This application indicated that even when fitting data over this frequency range, parameter estimates are not well defined and the parameter estimated with least accuracy is airway resistance.     

基于矩估计的遗传算法杂波幅度模型参数估计

选用了适用于多参数估计的遗传算法对Ku波段实测杂波数据进行K分布以及对数正态分布模型进行参数估计.为减少搜索范围,采用了计算量小的矩估计方法确定参数初始值;以常用的最小均方误差为目标函数.结果显示在遗传代数为50左右适应度函数收敛到稳定值,且拟合曲线与杂波数据直方图很好吻合.验证了遗传算法用于杂波模型参数估计的适用性.     

The use of residuals in statistical modeling for fatigue-crackgrowth under constant amplitude loading

Various models have been proposed over the years to fit crack growth data. Many papers have appeared in which one or more models are mooted and fitted, and various assessments made of the quality of the fit. At the basic level the data are plotted, together with the fitted curve, to show agreement of experimental and predicted values. In this paper we suggest that it can be useful and informative to go one step further, examining the residuals, i.e., the differences between the experimental and predicted values. We draw attention to certain statistical methods for such critical assessment and show by example that this can reveal deficiencies in fit not otherwise obvious. In this way suitable modifications to the model can be suggested. Additional plots of estimated parameters are also shown to be informative about models     

Impact of ambient temperature set point deviation on Arrhenius estimates

A least squares model is developed that takes ambient temperature set point error into account when making predictions with the Arrhenius equation. Complete data following a lognormal distribution is assumed. According to the model, the regression can be performed assuming the temperatures are at their desired levels (fixed temperature case). Random variation of the actual ambient temperature about the set point value will inflate the variance or mean squared error but will not bias the estimates. This increases the width of confidence intervals on the parameter estimates and predictions like the MTTF compared to the fixed temperature case. The amount of inflated variance depends chiefly on the extent of set point deviation and activation energy but is also influenced by choice of experimental design.     

A high precision global prediction approach based on local prediction approaches

Traditional model-free prediction approaches, such as neural networks or fuzzy models use all training data without preference in building their prediction models. Alternately, one may make predictions based only on a set of the most recent data without using other data. Usually, such local prediction schemes may have better performance in predicting time series than global prediction schemes do. However, local prediction schemes only use the most recent information and ignore information bearing on far away data. As a result, the accuracy of local prediction schemes may be limited. In this paper a novel prediction approach, termed the Markov-Fourier gray model (MFGM), is proposed. The approach builds a gray model from a set of the most recent data and a Fourier series is used to fit the residuals produced by this gray model. Then, the Markov matrices are employed to encode possible global information generated also by the residuals. It is evident that MFGM can provide the best performance among existing prediction schemes. Besides, we also implemented a short-term MFGM approach, in which the Markov matrices only recorded information for a period of time instead of all data. The predictions using MFGM again are more accurate than those using short-term MFGM. Thus, it is concluded that the global information encoded in the Markov matrices indeed can provide useful information for predictions.     

Parameter estimation and screening of solar cells

The aggregation (sorting) of the individual solar cells into an array is commonly based on a single operating point on the current-voltage (I-V) characteristic curve. an alternative approach for cell performance prediction and cell screening is provided by modelling the cell using an equivalent electrical circuit, in which the parameters involved are related to the physical phenomena in the device. These analytical models may be represented by a double exponential I-V characteristic with seven parameters, by a double exponential model with five parameters or by a single exponential equation with four or five parameters. In this article we address issues concerning methodologies for the determination of solar cell parameters based on measured data points of the I-V characteristic, and introduce a procedure for screening solar cells for arrays. We show that common curve-fitting techniques, e.g. least-squares, may produce many combinations of parameter values while maintaining a good fit between the fitted and measured I-V characteristics of the cell. Therefore, techniques relying on curve-fitting criteria alone cannot be used directly for cell parameterization. We propose a consistent procedure that takes into account the entire set of parameter values for a batch of cells. This procedure is based on a definition of a mean cell representing the batch, and takes into account the relative contribution of each parameter to the overall goodness of fit. the procedure is demonstrated on a batch of 50 silicon cells for Space Station Freedom.     

Influence of flow pattern on the parameter estimates of a simplebreathing mechanics model

The first-order model of breathing mechanics is widely used in clinical practice to assess the viscoelastic properties of the respiratory system. Although simple, this model takes the predominant features of the pressure-flow relationship into account but gives highly systematic residuals between measured and model-predicted variables. To achieve a better fit of the entire data set, an approach hypothesizing deterministic time-variations of model parameters, summarized by information-weighted histograms was recently proposed by Bates and Lauzon (1992). The present study uses flow and pressure data measured in intensive care patients to evaluate the real potential of this approach in clinical practice. Information-weighted histograms of the model parameters, estimated by an on-line identification algorithm, were first constructed by taking into account the parameter percentage standard deviations. Then, the influence of the respiratory flow pattern on the calculated histograms was evaluated by the Kolmogorov-Smirnov statistical test. The results show that the method gives good reproducibility under stable experimental conditions. In addition, for a given airflow waveform; an increase in respiratory frequency shifts the histograms representing time-varying viscous properties strongly versus lower values, whereas it shifts the histograms representing time-varying elastic properties slightly versus higher values. On the other hand, the same histograms were highly dependent on the airflow waveform, especially for the viscous properties. Even in a limited experimental work, in all the conditions considered, the method provides results which agree well with the physiological knowledge of nonlinear and multicompartment behavior of respiratory mechanics     

Wavelet-based estimation of 1/f-type signal parameters: confidenceintervals using the bootstrap

We propose to construct confidence intervals of parameters of 1/f-type signals using a nonparametric wavelet-based bootstrap method. Bootstrap-based confidence intervals of maximum likelihood parameter estimates are compared to the confidence intervals derived from the Cramer-Rao lower bound (CRLB). For moderately large data sample sizes, the bootstrap approach achieves the nominal coverage and may perform better than the CRLB-based parametric approach     

Calculation of average coding efficiency based on subjective quality scores

The Bjøntegaard model is widely used to calculate the coding efficiency between different codecs. However, this model might not be an accurate predictor of the true coding efficiency as it relies on PSNR measurements. Therefore, in this paper, we propose a model to calculate the average coding efficiency based on subjective quality scores, i.e., mean opinion scores (MOS). We call this approach Subjective Comparison of ENcoders based on fItted Curves (SCENIC). To consider the intrinsic nature of bounded rating scales, a logistic function is used to fit the rate–distortion (R–D) values. The average MOS and bit rate differences are computed between the fitted R–D curves. The statistical property of subjective scores is considered to estimate corresponding confidence intervals on the calculated average MOS and bit rate differences. The proposed model is expected to report more realistic coding efficiency as PSNR is not always correlated with perceived visual quality.     

Dynamic HIV/AIDS parameter estimation with application to a vaccine readiness study in southern Africa

This paper proposes a procedure of parameter estimation for all parameters of the three-dimensional HIV model. The least square based procedure uses standard optimization routines to allow parameter extraction for individual patients. It is shown how additional information from outside a measurement dataset can be included in the estimation routine to increase the reliability and accuracy of parameter estimates. A dataset from 44 patients of Southern Africa is analyzed to find the set point and the time until set point for these patients together with an estimate of the model parameters with confidence intervals for the cohort. The procedure is also applied to a long-term dataset of the HIV/AIDS progression to find possible variations in parameters.     

Some Graphical Techniques for Estimating Weibull Confidence Intervals

Many methods for estimating the parameter and percentile statistical confidence intervals for the Weibull and Gumbel (extreme value) distributions have been described in the literature. Most of these methods depend on extensive computer programs, require reference to tables which do not cover all sample sizes of interest and/or are not widely available. This paper describes a semi-empirical technique which permits rapid estimation of the 2-sided 90% statistical confidence intervals for the Weibull or Gumbel distribution parameters, as well as for the 1, 5, 10 percentiles. The estimates can be obtained for type II censoring and sample sizes to 25. The statistical confidence intervals calculated using this method are not exact, but are very good approximations and are useful to engineers who do not have ready access to programs or lengthy tables, or who require quick estimates. If more accurate statistical confidence intervals are required, then the more complicated methods described in the references should be used.     

基于稀疏性非负矩阵分解的偏振图像快速融合方法

针对基于非负矩阵分解(non-negative matrix factorization, NMF)的偏振图像融合方法效率低的不足,提出一种基于稀疏性NMF的偏振图像快速融合方法。首先,以偏振信息解析得到的各偏振参量图像构造原始数据集,其次,对NMF增加稀疏性约束,利用稀疏表示下的在线字典学习算法进行快速分解,然后对分解得到的三幅特征基图像按清晰度和方差进行排序,将排序后的特征基图像经直方图匹配及HSI颜色映射后,变换到RGB颜色空间,得到融合图像。与基于NMF的方法相比,运行时间提高约120倍,达到约1.5 s完成一次融合过程。实验结果验证了该方法在改善融合效果的同时,运行效率明显提高。     

Linear discriminant analysis in network traffic modelling

It is difficult to give an accurate judgement of whether the traffic model fit the actual traffic. The traditional method is to compare the Hurst parameter, data histogram and autocorrelation function. The method of comparing Hurst parameter cannot give exact results and judgement. The method of comparing data histogram and autocorrelation only gives a qualitative judgement. Based on linear discriminant analysis we proposed a novel arithmetic. Utilizing this arithmetic we analysed some sets of data with large and little differences. We also analysed some sets of data generated by network simulator. The analysis result is accurate. Comparing with traditional method, this arithmetic is useful and can conveniently give an accurate judgement for complex network traffic trace. Copyright © 2005 John Wiley & Sons, Ltd.     

System-reliability confidence-intervals for complex-systems withestimated component-reliability

A flexible procedure is described and demonstrated to determine approximate confidence intervals for system reliability when there is uncertainty regarding component reliability information. The approach is robust, and applies to many system-design configurations and component time-to-failure distributions, resulting in few restrictions for the use of these confidence intervals. The methods do not require any parametric assumptions for component reliability or time-to-failure, and allows type-I or -II censored data records. The confidence intervals are based on the variance of the component and system reliability estimates and a lognormal distribution assumption for the system reliability estimate. This approach applies to any system design which can be decomposed into series and/or parallel connections between the components. To evaluate the validity of the confidence limits, numerous simulations were performed for two hypothetical systems with different data sample-sizes and confidence levels. The test cases and empirical results demonstrate that this new method for estimating confidence intervals provides good coverage, can be readily applied, requires only minimal computational effort, and applies for a much greater range of design configurations and data types compared to other methods. For many design problems, these confidence intervals are preferable because there is no requirement for an exponential time-to-failure distribution nor are component data limited to binomial data     

Parameter estimation in stochastic mammogram model by heuristic optimization techniques.

The appearance of disproportionately large amounts of high-density breast parenchyma in mammograms has been found to be a strong indicator of the risk of developing breast cancer. Hence, the breast density model is popular for risk estimation or for monitoring breast density change in prevention or intervention programs. However, the efficiency of such a stochastic model depends on the accuracy of estimation of the model's parameter set. We propose a new approach-heuristic optimization-to estimate more accurately the model parameter set as compared to the conventional and popular expectation-maximization (EM) algorithm. After initial segmentation of a given mammogram, the finite generalized Gaussian mixture (FGGM) model is constructed by computing the statistics associated with different image regions. The model parameter set thus obtained is estimated by particle swarm optimization (PSO) and evolutionary programming (EP) techniques, where the objective function to be minimized is the relative entropy between the image histogram and the estimated density distributions. When our heuristic approach was applied to different categories of mammograms from the Mini-MIAS database, it yielded lower floor of estimation error in 109 out of 112 cases (97.3 %), and 101 out of 102 cases (99.0%), for the number of image regions being five and eight, respectively, with the added advantage of faster convergence rate, when compared to the EM approach. Besides, the estimated density model preserves the number of regions specified by the information-theoretic criteria in all the test cases, and the assessment of the segmentation results by radiologists is promising.     

Systematic autoregulation counteracts the carotid baroreflex

The interaction between autoregulation and baroregulation and its effect on the gains of the short-term pressure regulatory system was studied by performing both open- and closed-loop experiments in the same five anesthetized, vagotomized dogs, and by analyzing the data making use of a new model. With carotid pressure constant (no baroregulation) the pressure-flow data were convex to the flow axis, thus indicating the presence of autoregulation. When baroregulation was present the data were convex to the pressure axis. Our model was able to fit the data as measured in both cases. From the fitting procedure the zero-flow pressure intercept Pzf, the autoregulation resistance gain Gra, and the baroregulation resistance gain Grb were estimated. Pzf was about 20 mmHg in three dogs and about zero in the other two. Average values of Gra and Grb were 13.0 +/- 3.5 mmHg min2/L2 and 0.83 +/- 0.25 min/L, respectively. The two curves which fitted the data points collected in the presence and in the absence of baroreflex intersected at a point (Qo, Po) generally different from the control point. We determined the open-loop gain, Goc = GrbQo, about the point (Qo, Po). The averaged value was 2.23 +/- 0.84. When autoregulation was neglected, the resistance gain Grb and the open-loop gain Goc obtained from the same closed-loop method were underestimated (0.32 +/- 0.15 min/L and 0.88 +/- 0.48, respectively). In the open-loop preparation the carotid sinuses were isolated and the aortic (P) versus carotid (Pca) pressure data were collected. A third-order polynomial was fitted to these data.(ABSTRACT TRUNCATED AT 250 WORDS)     

Simultaneous confidence intervals for all ratios to the best: theexponential distribution

Consider k independent exponential populations with different scale and location (possibly unknown) parameters. A set of simultaneous upper confidence intervals for all ratios to the largest scale parameter is derived. The data are assumed to be: (1) complete; or (2) incomplete with type-II censoring. The cases of known and unknown location parameters are treated separately     

Physiological interpretations based on lumped element models fit torespiratory impedance data: use of forward-inverse modeling

Respiratory impedance (Zrs) data at lower (less than 4 Hz) and higher (greater than 32 Hz) frequencies require more complicated inverse models than the standard series combination of a respiratory resistance, inertance, and compliance. In this paper, a forward-inverse modeling approach was used to provide insight on how the parameters in these more complicated inverse models reflect the true physiological system. Forward models are set up to incorporate explicit physiological and anatomical detail. Simulated forward data are then fit with identifiable inverse models and the parameter estimates related to the known detail in the forward model. It is shown that inverse fitting of low frequency data alone will not allow a distinction between frequency dependence due to airway inhomogeneities and frequency dependence due to tissue viscoelasticity. With higher frequency data, a forward model based on an asymmetric branching airways network was used to simulate Zrs from 0.1-128 Hz with increasing amounts of nonuniform peripheral airway obstruction. Here, inverse modeling is more amenable to sensibly separating estimates of airway and tissue properties. A key result, however, is that changes in the tissue parameters of an inverse model (which provides an excellent fit to Zrs data) will appropriately occur in response to inhomogeneous alterations in airway diameters only. The apparent altered tissue properties reflect the decreased communication of some tissue segments with the airway opening and not an explicit change at the tissue level. These phenomena present a substantial problem for the inverse modeler. Finally, inverse model fitting of low and high frequency Zrs data simultaneously with a single model is not helpful for extracting additional physiological detail. Instead, separate models should be applied to each frequency range.     

A model for residual life prediction based on Brownian motion with an adaptive drift

A degradation model is presented in this paper for the prediction of the residual life using an adapted Brownian motion-based approach with a drifting parameter. This model differs from other Brownian motion-based approaches in that the drifting parameter of the degradation process is adapted to the history of monitored information. This adaptation is performed by Kalman filtering. We also use a threshold distribution instead of the usual single threshold line which is sometime difficult to obtain in practice. We demonstrate the model using some examples and show that the model performs reasonably well and has a better prediction ability than the standard Brownian motion-based model. The model is then fitted to the data generated from a simulator using the expectation-maximization algorithm. We also fit a standard Brownian motion-based model to the same data to compare the difference and performance. The result shows that the adapted model performs better in terms of certain test statistics and the total mean square errors.     

A lucky drift model,including a soft threshold energy,fitted to experimental measurements of ionization coefficients

We have fitted the lucky drift model of impact ionization due to Burt to experimental data for GaAs, InP, Si and In_0.53Ga_0.47As. The agreement is surprisingly good considering the simplicity of that model. The lucky drift model has been further extended to include the effects of a soft threshold energy where carriers do not necessarily ionize immediately on achieving the threshold energy. This new lucky drift model is found to fit the same experimental data better than the model due to Burt in all cases but one.