Search results for: likelihood method

Authors: M. Qamarul Islam, Ergun Dogan, Mehmet Yazici

Abstract:

There is a growing body of evidence that non-normal data is more prevalent in nature than the normal one. Examples can be quoted from, but not restricted to, the areas of Economics, Finance and Actuarial Science. The non-normality considered here is expressed in terms of fat-tailedness and asymmetry of the relevant distribution. In this study a skew t distribution that can be used to model a data that exhibit inherent non-normal behavior is considered. This distribution has tails fatter than a normal distribution and it also exhibits skewness. Although maximum likelihood estimates can be obtained by solving iteratively the likelihood equations that are non-linear in form, this can be problematic in terms of convergence and in many other respects as well. Therefore, it is preferred to use the method of modified maximum likelihood in which the likelihood estimates are derived by expressing the intractable non-linear likelihood equations in terms of standardized ordered variates and replacing the intractable terms by their linear approximations obtained from the first two terms of a Taylor series expansion about the quantiles of the distribution. These estimates, called modified maximum likelihood estimates, are obtained in closed form. Hence, they are easy to compute and to manipulate analytically. In fact the modified maximum likelihood estimates are equivalent to maximum likelihood estimates, asymptotically. Even in small samples the modified maximum likelihood estimates are found to be approximately the same as maximum likelihood estimates that are obtained iteratively. It is shown in this study that the modified maximum likelihood estimates are not only unbiased but substantially more efficient than the commonly used moment estimates or the least square estimates that are known to be biased and inefficient in such cases. Furthermore, in conventional regression analysis, it is assumed that the error terms are distributed normally and, hence, the well-known least square method is considered to be a suitable and preferred method for making the relevant statistical inferences. However, a number of empirical researches have shown that non-normal errors are more prevalent. Even transforming and/or filtering techniques may not produce normally distributed residuals. Here, a study is done for multiple linear regression models with random error having non-normal pattern. Through an extensive simulation it is shown that the modified maximum likelihood estimates of regression parameters are plausibly robust to the distributional assumptions and to various data anomalies as compared to the widely used least square estimates. Relevant tests of hypothesis are developed and are explored for desirable properties in terms of their size and power. The tests based upon modified maximum likelihood estimates are found to be substantially more powerful than the tests based upon least square estimates. Several examples are provided from the areas of Economics and Finance where such distributions are interpretable in terms of efficient market hypothesis with respect to asset pricing, portfolio selection, risk measurement and capital allocation, etc.

Keywords: least square estimates, linear regression, maximum likelihood estimates, modified maximum likelihood method, non-normality, robustness

Procedia PDF Downloads 375

Authors: Kajingulu Malandala, Ranganai Edmore

Abstract:

The asymmetric Laplace distribution (ADL) is commonly used as the likelihood function of the Bayesian quantile regression, and it offers different families of likelihood method for quantile regression. Notwithstanding their popularity and practicality, ADL is not smooth and thus making it difficult to maximize its likelihood. Furthermore, Bayesian inference is time consuming and the selection of likelihood may mislead the inference, as the Bayes theorem does not automatically establish the posterior inference. Furthermore, ADL does not account for greater skewness and Kurtosis. This paper develops a new aspect of quantile regression approach for count data based on inverse of the cumulative density function of the Poisson, binomial and Delaporte distributions using the integrated nested Laplace Approximations. Our result validates the benefit of using the integrated nested Laplace Approximations and support the approach for count data.

Keywords: quantile regression, Delaporte distribution, count data, integrated nested Laplace approximation

Procedia PDF Downloads 128

Authors: I. V. Pinto, M. R. Sooriyarachchi

Abstract:

It can be frequently observed that the data arising in our environment have a hierarchical or a nested structure attached with the data. Multilevel modelling is a modern approach to handle this kind of data. When multilevel modelling is combined with a binary response, the estimation methods get complex in nature and the usual techniques are derived from quasi-likelihood method. The estimation methods which are compared in this study are, marginal quasi-likelihood (order 1 & order 2) (MQL1, MQL2) and penalized quasi-likelihood (order 1 & order 2) (PQL1, PQL2). A statistical model is of no use if it does not reflect the given dataset. Therefore, checking the adequacy of the fitted model through a goodness-of-fit (GOF) test is an essential stage in any modelling procedure. However, prior to usage, it is also equally important to confirm that the GOF test performs well and is suitable for the given model. This study assesses the suitability of the GOF test developed for binary response multilevel models with respect to the method used in model estimation. An extensive set of simulations was conducted using MLwiN (v 2.19) with varying number of clusters, cluster sizes and intra cluster correlations. The test maintained the desirable Type-I error for models estimated using PQL2 and it failed for almost all the combinations of MQL. Power of the test was adequate for most of the combinations in all estimation methods except MQL1. Moreover, models were fitted using the four methods to a real-life dataset and performance of the test was compared for each model.

Keywords: goodness-of-fit test, marginal quasi-likelihood, multilevel modelling, penalized quasi-likelihood, power, quasi-likelihood, type-I error

Procedia PDF Downloads 115

Authors: Wikanda Phaphan

Abstract:

Mixture generalized gamma distribution is a combination of two distributions: generalized gamma distribution and length biased generalized gamma distribution. These two distributions were presented by Suksaengrakcharoen and Bodhisuwan in 2014. The findings showed that probability density function (pdf) had fairly complexities, so it made problems in estimating parameters. The problem occurred in parameter estimation was that we were unable to calculate estimators in the form of critical expression. Thus, we will use numerical estimation to find the estimators. In this study, we presented a new method of the parameter estimation by using the expectation – maximization algorithm (EM), the conjugate gradient method, and the quasi-Newton method. The data was generated by acceptance-rejection method which is used for estimating α, β, λ and p. λ is the scale parameter, p is the weight parameter, α and β are the shape parameters. We will use Monte Carlo technique to find the estimator's performance. Determining the size of sample equals 10, 30, 100; the simulations were repeated 20 times in each case. We evaluated the effectiveness of the estimators which was introduced by considering values of the mean squared errors and the bias. The findings revealed that the EM-algorithm had proximity to the actual values determined. Also, the maximum likelihood estimators via the conjugate gradient and the quasi-Newton method are less precision than the maximum likelihood estimators via the EM-algorithm.

Keywords: conjugate gradient method, quasi-Newton method, EM-algorithm, generalized gamma distribution, length biased generalized gamma distribution, maximum likelihood method

Procedia PDF Downloads 196

Authors: Autcha Araveeporn

Abstract:

This paper is to compare the parameter estimation of the mean in normal distribution by Maximum Likelihood (ML), Bayes, and Markov Chain Monte Carlo (MCMC) methods. The ML estimator is estimated by the average of data, the Bayes method is considered from the prior distribution to estimate Bayes estimator, and MCMC estimator is approximated by Gibbs sampling from posterior distribution. These methods are also to estimate a parameter then the hypothesis testing is used to check a robustness of the estimators. Data are simulated from normal distribution with the true parameter of mean 2, and variance 4, 9, and 16 when the sample sizes is set as 10, 20, 30, and 50. From the results, it can be seen that the estimation of MLE, and MCMC are perceivably different from the true parameter when the sample size is 10 and 20 with variance 16. Furthermore, the Bayes estimator is estimated from the prior distribution when mean is 1, and variance is 12 which showed the significant difference in mean with variance 9 at the sample size 10 and 20.

Keywords: Bayes method, Markov chain Monte Carlo method, maximum likelihood method, normal distribution

Procedia PDF Downloads 326

Authors: Khandoker Akib Mohammad

Abstract:

While analyzing time-to-event data, it is possible that a certain fraction of subjects will never experience the event of interest, and they are said to be cured. When this feature of survival models is taken into account, the models are commonly referred to as cure models. In the presence of covariates, the conditional survival function of the population can be modelled by using the cure model, which depends on the probability of being uncured (incidence) and the conditional survival function of the uncured subjects (latency), and a combination of logistic regression and Cox proportional hazards (PH) regression is used to model the incidence and latency respectively. In this paper, we have shown the asymptotic normality of the profile likelihood estimator via asymptotic expansion of the profile likelihood and obtain the explicit form of the variance estimator with an implicit function in the profile likelihood. We have also shown the efficient score function based on projection theory and the profile likelihood score function are equal. Our contribution in this paper is that we have expressed the efficient information matrix as the variance of the profile likelihood score function. A simulation study suggests that the estimated standard errors from bootstrap samples (SMCURE package) and the profile likelihood score function (our approach) are providing similar and comparable results. The numerical result of our proposed method is also shown by using the melanoma data from SMCURE R-package, and we compare the results with the output obtained from the SMCURE package.

Keywords: Cox PH model, cure model, efficient score function, EM algorithm, implicit function, profile likelihood

Procedia PDF Downloads 107

Authors: Al Omari Moahmmed Ahmed

Abstract:

These papers describe the Bayesian Estimator using Markov Chain Monte Carlo and Lindley’s approximation and the maximum likelihood estimation of the Weibull distribution with Type-I censored data. The maximum likelihood method can’t estimate the shape parameter in closed forms, although it can be solved by numerical methods. Moreover, the Bayesian estimates of the parameters, the survival and hazard functions cannot be solved analytically. Hence Markov Chain Monte Carlo method and Lindley’s approximation are used, where the full conditional distribution for the parameters of Weibull distribution are obtained via Gibbs sampling and Metropolis-Hastings algorithm (HM) followed by estimate the survival and hazard functions. The methods are compared to Maximum Likelihood counterparts and the comparisons are made with respect to the Mean Square Error (MSE) and absolute bias to determine the better method in scale and shape parameters, the survival and hazard functions.

Keywords: weibull distribution, bayesian method, markov chain mote carlo, survival and hazard functions

Procedia PDF Downloads 448

Authors: Mesut BALIBEY, Serpil TÜRKYILMAZ

Abstract:

A popular topic in the econometrics and time series area is the cointegrating relationships among the components of a nonstationary time series. Engle and Granger’s least squares method and Johansen’s conditional maximum likelihood method are the most widely-used methods to determine the relationships among variables. Furthermore, a method proposed to test a unit root based on the periodogram ordinates has certain advantages over conventional tests. Periodograms can be calculated without any model specification and the exact distribution under the assumption of a unit root is obtained. For higher order processes the distribution remains the same asymptotically. In this study, in order to indicate advantages over conventional test of periodograms, we are going to examine a possible relationship between tourism and economic growth during the period 1999:01-2010:12 for Turkey by using periodogram method, Johansen’s conditional maximum likelihood method, Engle and Granger’s ordinary least square method.

Keywords: cointegration, economic growth, periodogram ordinate, tourism

Procedia PDF Downloads 238

Authors: Yi-Cheng Tian, Miin-Shen Yang

Abstract:

The mixture likelihood approach to clustering is a popular clustering method where the expectation and maximization (EM) algorithm is the most used mixture likelihood method. In the literature, the EM algorithm had been used for mixture regression models. However, these EM mixture regression algorithms are sensitive to initial values with a priori number of clusters. In this paper, to resolve these drawbacks, we construct a learning-based schema for the EM mixture regression algorithm such that it is free of initializations and can automatically obtain an approximately optimal number of clusters. Some numerical examples and comparisons demonstrate the superiority and usefulness of the proposed learning-based EM mixture regression algorithm.

Keywords: clustering, EM algorithm, Gaussian mixture model, mixture regression model

Procedia PDF Downloads 475

Authors: Al Omari Moahmmed Ahmed

Abstract:

In the Bayesian, we developed an approach by using non-informative prior with covariate and obtained by using Gauss quadrature method to estimate the parameters of the covariate and reliability function of the Weibull regression distribution with Type-I censored data. The maximum likelihood seen that the estimators obtained are not available in closed forms, although they can be solved it by using Newton-Raphson methods. The comparison criteria are the MSE and the performance of these estimates are assessed using simulation considering various sample size, several specific values of shape parameter. The results show that Bayesian with non-informative prior is better than Maximum Likelihood Estimator.

Keywords: non-informative prior, Bayesian method, type-I censoring, Gauss quardature

Procedia PDF Downloads 465

Authors: Tasir Khan, Yejuan Wan, Kalim Ullah

Abstract:

Hydrological frequency analysis has been conducted to estimate the minimum flow elevation of the Indus River in Pakistan to protect the ecosystem. The Maximum likelihood estimation (MLE) technique is used to estimate the best-fitted distribution for Minimum Ecological Flows at nine stations of the Indus River in Pakistan. The four selected distributions, Generalized Extreme Value (GEV) distribution, Generalized Logistics (GLO) distribution, Generalized Pareto (GPA) distribution, and Pearson type 3 (PE3) are fitted in all sites, usually used in hydro frequency analysis. Compare the performance of these distributions by using the goodness of fit tests, such as the Kolmogorov Smirnov test, Anderson darling test, and chi-square test. The study concludes that the Maximum Likelihood Estimation (MLE) method recommended that GEV and GPA are the most suitable distributions which can be effectively applied to all the proposed sites. The quantiles are estimated for the return periods from 5 to 1000 years by using MLE, estimations methods. The MLE is the robust method for larger sample sizes. The results of these analyses can be used for water resources research, including water quality management, designing irrigation systems, determining downstream flow requirements for hydropower, and the impact of long-term drought on the country's aquatic system.

Keywords: minimum ecological flow, frequency distribution, indus river, maximum likelihood estimation

Procedia PDF Downloads 51

Authors: Ilaria Lucrezia Amerise

Abstract:

The statistical modelling of precipitation data for a given portion of territory is fundamental for the monitoring of climatic conditions and for Hydrogeological Management Plans (HMP). This modelling is rendered particularly complex by the changes taking place in the frequency and intensity of precipitation, presumably to be attributed to the global climate change. This paper applies the Wakeby distribution (with 5 parameters) as a theoretical reference model. The number and the quality of the parameters indicate that this distribution may be the appropriate choice for the interpolations of the hydrological variables and, moreover, the Wakeby is particularly suitable for describing phenomena producing heavy tails. The proposed estimation methods for determining the value of the Wakeby parameters are the same as those used for density functions with heavy tails. The commonly used procedure is the classic method of moments weighed with probabilities (probability weighted moments, PWM) although this has often shown difficulty of convergence, or rather, convergence to a configuration of inappropriate parameters. In this paper, we analyze the problem of the likelihood estimation of a random variable expressed through its quantile function. The method of maximum likelihood, in this case, is more demanding than in the situations of more usual estimation. The reasons for this lie, in the sampling and asymptotic properties of the estimators of maximum likelihood which improve the estimates obtained with indications of their variability and, therefore, their accuracy and reliability. These features are highly appreciated in contexts where poor decisions, attributable to an inefficient or incomplete information base, can cause serious damages.

Keywords: generalized extreme values, likelihood estimation, precipitation data, Wakeby distribution

Procedia PDF Downloads 103

Authors: Hao Chen, Bo Guo, Ping Jiang

Abstract:

Residual life (RL) estimation based on multi-phase nonlinear Wiener process was studied in this paper, which is significant for complicated products with small samples. Firstly, nonlinear Wiener model with random parameter was introduced and multi-phase nonlinear Wiener model was proposed to model degradation process of products that were nonlinear and separated into different phases. Then the multi-phase RL probability density function based on the presented model was derived approximately in a closed form and parameters estimation was achieved with the method of maximum likelihood estimation (MLE). Finally, the method was applied to estimate the RL of high voltage plus capacitor. Compared with the other three different models by log-likelihood function (Log-LF) and Akaike information criterion (AIC), the results show that the proposed degradation model can capture degradation process of high voltage plus capacitors in a better way and provide a more reliable result.

Keywords: multi-phase nonlinear wiener process, residual life estimation, maximum likelihood estimation, high voltage plus capacitor

Procedia PDF Downloads 423

Authors: Dewi Retno Sari Saputro, Purnami Widyaningsih, Hendrika Handayani

Abstract:

Extreme data on an observation can occur due to unusual circumstances in the observation. The data can provide important information that can’t be provided by other data so that its existence needs to be further investigated. The method for obtaining extreme data is one of them using maxima block method. The distribution of extreme data sets taken with the maxima block method is called the distribution of extreme values. Distribution of extreme values is Gumbel distribution with two parameters. The parameter estimation of Gumbel distribution with maximum likelihood method (ML) is difficult to determine its exact value so that it is necessary to solve the approach. The purpose of this study was to determine the parameter estimation of Gumbel distribution with quasi-Newton BFGS method. The quasi-Newton BFGS method is a numerical method used for nonlinear function optimization without constraint so that the method can be used for parameter estimation from Gumbel distribution whose distribution function is in the form of exponential doubel function. The quasi-New BFGS method is a development of the Newton method. The Newton method uses the second derivative to calculate the parameter value changes on each iteration. Newton's method is then modified with the addition of a step length to provide a guarantee of convergence when the second derivative requires complex calculations. In the quasi-Newton BFGS method, Newton's method is modified by updating both derivatives on each iteration. The parameter estimation of the Gumbel distribution by a numerical approach using the quasi-Newton BFGS method is done by calculating the parameter values that make the distribution function maximum. In this method, we need gradient vector and hessian matrix. This research is a theory research and application by studying several journals and textbooks. The results of this study obtained the quasi-Newton BFGS algorithm and estimation of Gumbel distribution parameters. The estimation method is then applied to daily rainfall data in Purworejo District to estimate the distribution parameters. This indicates that the high rainfall that occurred in Purworejo District decreased its intensity and the range of rainfall that occurred decreased.

Keywords: parameter estimation, Gumbel distribution, maximum likelihood, broyden fletcher goldfarb shanno (BFGS)quasi newton

Procedia PDF Downloads 295

Authors: Al Omari Mohammed Ahmed

Abstract:

This paper focuses on Maximum Likelihood Estimator with Covariate. Covariates are incorporated into the Weibull model. Under this regression model with regards to maximum likelihood estimator, the parameters of the covariate, shape parameter, survival function and hazard rate of the Weibull regression distribution with right censored data are estimated. The mean square error (MSE) and absolute bias are used to compare the performance of Weibull regression distribution. For the simulation comparison, the study used various sample sizes and several specific values of the Weibull shape parameter.

Keywords: weibull regression distribution, maximum likelihood estimator, survival function, hazard rate, right censoring

Procedia PDF Downloads 411

Authors: Peter A. L’vov, Roman S. Konovalov, Alexey A. L’vov

Abstract:

The paper presents the advanced digital modification of the conventional current loop circuit for pressure piezoelectric transducers. The optimal DSP algorithms of current loop responses by the maximum likelihood method are applied for diminishing of measurement errors. The loop circuit has some additional advantages such as the possibility to operate with any type of resistance or reactance sensors, and a considerable increase in accuracy and quality of measurements to be compared with AC bridges. The results obtained are dedicated to replace high-accuracy and expensive measuring bridges with current loop circuits.

Keywords: current loop, maximum likelihood method, optimal digital signal processing, precise pressure measurement

Procedia PDF Downloads 498

Authors: Mahdy ‎Esmailian, Mahdi ‎Doostparast, Ahmad ‎Parsian

Abstract:

‎In this article‎, ‎sequential order statistics (SOS) censoring type II samples coming from the generalized Pareto distribution are considered‎. ‎Maximum likelihood (ML) estimators of the unknown parameters are derived on the basis of the available multiple SOS data‎. ‎Necessary conditions for existence and uniqueness of the derived ML estimates are given‎. Due to complexity in the proposed likelihood function‎, ‎a useful re-parametrization is suggested‎. ‎For illustrative purposes‎, ‎a Monte Carlo simulation study is conducted and an illustrative example is analysed‎.

Keywords: bayesian estimation‎, generalized pareto distribution‎, ‎maximum likelihood estimation‎, sequential order statistics

Procedia PDF Downloads 469

Authors: Gamze Ozel, Selen Cakmakyapan

Abstract:

We introduce a new model called the Marshall-Olkin Rayleigh distribution which extends the Rayleigh distribution using Marshall-Olkin transformation and has increasing and decreasing shapes for the hazard rate function. Various structural properties of the new distribution are derived including explicit expressions for the moments, generating and quantile function, some entropy measures, and order statistics are presented. The model parameters are estimated by the method of maximum likelihood and the observed information matrix is determined. The potentiality of the new model is illustrated by means of real life data set.

Keywords: Marshall-Olkin distribution, Rayleigh distribution, estimation, maximum likelihood

Procedia PDF Downloads 461

Authors: Abd El Hady N. Ebraheim

Abstract:

This paper introduces a new generalization of the two parameter Weibull distribution. To this end, the quadratic rank transmutation map has been used. This new distribution is named exponentiated transmuted Weibull (ETW) distribution. The ETW distribution has the advantage of being capable of modeling various shapes of aging and failure criteria. Furthermore, eleven lifetime distributions such as the Weibull, exponentiated Weibull, Rayleigh and exponential distributions, among others follow as special cases. The properties of the new model are discussed and the maximum likelihood estimation is used to estimate the parameters. Explicit expressions are derived for the quantiles. The moments of the distribution are derived, and the order statistics are examined.

Keywords: exponentiated, inversion method, maximum likelihood estimation, transmutation map

Procedia PDF Downloads 537

Authors: K. A. Adepoju, O. I. Shittu, A. U. Chukwu

Abstract:

In this paper, a new four-parameter generalized version of the Fisher Snedecor distribution called Beta- F distribution is introduced. The comprehensive account of the statistical properties of the new distributions was considered. Formal expressions for the cumulative density function, moments, moment generating function and maximum likelihood estimation, as well as its Fisher information, were obtained. The flexibility of this distribution as well as its robustness using cancer remission time data was demonstrated. The new distribution can be used in most applications where the assumption underlying the use of other lifetime distributions is violated.

Keywords: fisher-snedecor distribution, beta-f distribution, outlier, maximum likelihood method

Procedia PDF Downloads 312

Authors: Guinevere Servis

Abstract:

This paper hypothesizes that withdrawal from positive self-awareness, and the increase of counterfactual-thinking and self-handicapping can help provide ample justification for an individual before, during and/or after committing a crime. The understanding of who someone is, one’s perspective on the world, and why they think the way they do is key to therapy in preventing recidivism. Developing habits to escape self-awareness, by using self-handicapping and counterfactual ideologies, may provide the necessary thinking patterns to decide disobeying the law is a worthy act to pursue. An increase in self-awareness is hypothesized to decrease the likelihood of recidivism, and ways of thinking that withdraw from self awareness can increase the likelihood of it. Especially for those who have been disadvantaged in life and disobeyed the law, self-handicapping and counterfactual thinking can also help to justify one's wrongdoing. The understanding of how a criminal views their disadvantages in the world, and one’s thinking patterns are hypothesized to help one better understand the entire scope on why a crime was committed and thus reduce the likelihood of recidivism. Utilizing therapy for prisoners to increase self-awareness of both thought and action can lead to a healthy, happier life and reduce the likelihood of reoffending. By discussing the terms associated with self-awareness theory and other psychological topics such as self-handicapping, counterfactual thinking, this paper argues the actions towards increasing positive self-awareness can help decrease likelihood of recidivism. Adversely, hypothesizing that increasing the ways of thinking that withdraw one from self-awareness, through counterfactual thinking and self-handicapping, can inherently increase the likelihood of recidivism. Evaluating these findings to further understand the needed changes in correctional institutions is fundamental to reducing crime, benefiting the criminal, the victim(s) and their family, and the state.

Keywords: crime, self-awareness theory, correctional institutions, self-regulate, counterfactual thinking, recidivism

Procedia PDF Downloads 51

Authors: Sami Mestiri, Abdeljelil Farhat

Abstract:

The aim of this current paper is to predict the credit risk of banks in Tunisia, over the period (2000-2005). For this purpose, two methods for the estimation of the logistic regression model with random effects: Penalized Quasi Likelihood (PQL) method and Gibbs Sampler algorithm are applied. By using the information on a sample of 528 Tunisian firms and 26 financial ratios, we show that Bayesian approach improves the quality of model predictions in terms of good classification as well as by the ROC curve result.

Keywords: forecasting, credit risk, Penalized Quasi Likelihood, Gibbs Sampler, logistic regression with random effects, curve ROC

Procedia PDF Downloads 508

Authors: Randa Alharbi, Vladislav Vyshemirsky

Abstract:

Systems biology is an important field in science which focuses on studying behaviour of biological systems. Modelling is required to produce detailed description of the elements of a biological system, their function, and their interactions. A well-designed model requires selecting a suitable mechanism which can capture the main features of the system, define the essential components of the system and represent an appropriate law that can define the interactions between its components. Complex biological systems exhibit stochastic behaviour. Thus, using probabilistic models are suitable to describe and analyse biological systems. Continuous-Time Markov Chain (CTMC) is one of the probabilistic models that describe the system as a set of discrete states with continuous time transitions between them. The system is then characterised by a set of probability distributions that describe the transition from one state to another at a given time. The evolution of these probabilities through time can be obtained by chemical master equation which is analytically intractable but it can be simulated. Uncertain parameters of such a model can be inferred using methods of Bayesian inference. Yet, inference in such a complex system is challenging as it requires the evaluation of the likelihood which is intractable in most cases. There are different statistical methods that allow simulating from the model despite intractability of the likelihood. Approximate Bayesian computation is a common approach for tackling inference which relies on simulation of the model to approximate the intractable likelihood. Particle Markov chain Monte Carlo (PMCMC) is another approach which is based on using sequential Monte Carlo to estimate intractable likelihood. However, both methods are computationally expensive. In this paper we discuss the efficiency and possible practical issues for each method, taking into account the computational time for these methods. We demonstrate likelihood-free inference by performing analysing a model of the Repressilator using both methods. Detailed investigation is performed to quantify the difference between these methods in terms of efficiency and computational cost.

Keywords: Approximate Bayesian computation(ABC), Continuous-Time Markov Chains, Sequential Monte Carlo, Particle Markov chain Monte Carlo (PMCMC)

Procedia PDF Downloads 176

Authors: Alvaro Callejas-Ramos, Lorena Alvarez-Perez, Alexander Benitez-Buenache, Anibal R. Figueiras-Vidal

Abstract:

This contribution presents a promising formulation which allows to extend the principled binary rebalancing procedures, also known as neutral re-balancing mechanisms in the sense that they do not alter the likelihood ratio

Keywords: Bregman divergences, imbalanced multiclass classifi-cation, informed re-balancing, invariant likelihood ratio

Procedia PDF Downloads 174

Authors: Naushad Mamode Khan

Abstract:

The Com-Poisson (CMP) model is one of the most popular discrete generalized linear models (GLMS) that handles both equi-, over- and under-dispersed data. In longitudinal context, an integer-valued autoregressive (INAR(1)) process that incorporates covariate specification has been developed to model longitudinal CMP counts. However, the joint likelihood CMP function is difficult to specify and thus restricts the likelihood based estimating methodology. The joint generalized quasilikelihood approach (GQL-I) was instead considered but is rather computationally intensive and may not even estimate the regression effects due to a complex and frequently ill conditioned covariance structure. This paper proposes a new GQL approach for estimating the regression parameters (GQLIII) that are based on a single score vector representation. The performance of GQL-III is compared with GQL-I and separate marginal GQLs (GQL-II) through some simulation experiments and is proved to yield equally efficient estimates as GQL-I and is far more computationally stable.

Keywords: longitudinal, com-Poisson, ill-conditioned, INAR(1), GLMS, GQL

Procedia PDF Downloads 327

Authors: M. Z. Raqab, Debasis Kundu, M. A. Meraou

Abstract:

In this paper, we have analyzed maximum precipitation data during a particular period of time obtained from different stations in the Global Historical Climatological Network of the USA. One important point to mention is that some stations are shut down on certain days for some reason or the other. Hence, the maximum values are recorded by excluding those readings. It is assumed that the number of stations that operate follows zero-truncated Poisson random variables, and the daily precipitation follows a lognormal random variable. We call this model a compound truncated Poisson lognormal model. The proposed model has three unknown parameters, and it can take a variety of shapes. The maximum likelihood estimators can be obtained quite conveniently using Expectation-Maximization (EM) algorithm. Approximate maximum likelihood estimators are also derived. The associated confidence intervals also can be obtained from the observed Fisher information matrix. Simulation results have been performed to check the performance of the EM algorithm, and it is observed that the EM algorithm works quite well in this case. When we analyze the precipitation data set using the proposed model, it is observed that the proposed model provides a better fit than some of the existing models.

Keywords: compound Poisson lognormal distribution, EM algorithm, maximum likelihood estimation, approximate maximum likelihood estimation, Fisher information, skew distribution

Procedia PDF Downloads 79

Authors: Daniel Fundi Murithi

Abstract:

Data from economic, social, clinical, and industrial studies are in some way incomplete or incorrect due to censoring. Such data may have adverse effects if used in the estimation problem. We propose the use of Maximum Likelihood Estimation (MLE) under a progressive type-II censoring scheme to remedy this problem. In particular, maximum likelihood estimates (MLEs) for the location (µ) and scale (λ) parameters of two Parameter Rayleigh distribution are realized under a progressive type-II censoring scheme using the Expectation-Maximization (EM) and the Newton-Raphson (NR) algorithms. These algorithms are used comparatively because they iteratively produce satisfactory results in the estimation problem. The progressively type-II censoring scheme is used because it allows the removal of test units before the termination of the experiment. Approximate asymptotic variances and confidence intervals for the location and scale parameters are derived/constructed. The efficiency of EM and the NR algorithms is compared given root mean squared error (RMSE), bias, and the coverage rate. The simulation study showed that in most sets of simulation cases, the estimates obtained using the Expectation-maximization algorithm had small biases, small variances, narrower/small confidence intervals width, and small root of mean squared error compared to those generated via the Newton-Raphson (NR) algorithm. Further, the analysis of a real-life data set (data from simple experimental trials) showed that the Expectation-Maximization (EM) algorithm performs better compared to Newton-Raphson (NR) algorithm in all simulation cases under the progressive type-II censoring scheme.

Keywords: expectation-maximization algorithm, maximum likelihood estimation, Newton-Raphson method, two-parameter Rayleigh distribution, progressive type-II censoring

Procedia PDF Downloads 126

Authors: Vandna Raghuvanshi, Vikrant Thakur, Anupam Jhobta

Abstract:

Objective: To design HRCT criterion to forecast the threat of pulmonary tuberculosis. Material and methods: This was a prospective study of 69 patients with clinical suspicion of pulmonary tuberculosis. We studied their medical characteristics, numerous separate HRCT-results, and a combination of HRCT findings to foresee the danger for PTB by utilizing univariate and multivariate investigation. Temporary HRCT diagnostic criteria were planned in view of these outcomes to find out the risk of PTB and tested these criteria on our patients. Results: The results of HRCT chest were analyzed, and Rank was given from 1 to 4 according to the HRCT chest findings. Sensitivity, specificity, positive predictive value, and negative predictive value were calculated. Rank 1: Highly suspected PTB. Rank 2: Probable PTB Rank 3: Nonspecific or difficult to differentiate from other diseases Rank 4: Other suspected diseases • Rank 1 (Highly suspected TB) was present in 22 (31.9%) patients, all of them finally diagnosed to have pulmonary tuberculosis. The sensitivity, specificity, and negative likelihood ratio for RANK 1 on HRCT chest was 53.6%, 100%, and 0.43, respectively. • Rank 2 (Probable TB) was present in 13 patients, out of which 12 were tubercular, and 1 was non-tubercular. • The sensitivity, specificity, positive likelihood ratio, and negative likelihood ratio of the combination of Rank 1 and Rank 2 was 82.9%, 96.4%, 23.22, and 0.18, respectively. • Rank 3 (Non-specific TB) was present in 25 patients, and out of these, 7 were tubercular, and 18 were non-tubercular. • When all these 3 ranks were considered together, the sensitivity approached 100% however, the specificity reduced to 35.7%. The positive likelihood ratio and negative likelihood ratio were 1.56 and 0, respectively. • Rank 4 (Other specific findings) was given to 9 patients, and all of these were non-tubercular. Conclusion: HRCT is useful in selecting individuals with greater chances of pulmonary tuberculosis.

Keywords: pulmonary, tuberculosis, multivariate, HRCT

Procedia PDF Downloads 137

Authors: Lanchi Xie, Zhihui Li, Zhigang Li, Guiqiang Wang, Lei Xu, Yuwen Yan

Abstract:

Image-based facial features can be classified into category recognition features and individual recognition features. Current automated face recognition systems extract a specific feature vector of different dimensions from a facial image according to their pre-trained neural network. However, to improve the efficiency of parameter calculation, an algorithm generally reduces the image details by pooling. The operation will overlook the details concerned much by forensic experts. In our experiment, we adopted a variety of face recognition algorithms based on deep learning, compared a large number of naturally collected face images with the known data of the same person's frontal ID photos. Downscaling and manual handling were performed on the testing images. The results supported that the facial recognition algorithms based on deep learning detected structural and morphological information and rarely focused on specific markers such as stains and moles. Overall performance, distribution of genuine scores and impostor scores, and likelihood ratios were tested to evaluate the accuracy of biometric systems and forensic experts. Experiments showed that the biometric systems were skilled in distinguishing category features, and forensic experts were better at discovering the individual features of human faces. In the proposed approach, a fusion was performed at the score level. At the specified false accept rate, the framework achieved a lower false reject rate. This paper contributes to improving the interpretability of the objective method of facial comparison and provides a novel method for human-machine collaboration in this field.

Keywords: likelihood ratio, automated facial recognition, facial comparison, biometrics

Procedia PDF Downloads 90

Authors: Ariful Islam, Showkat Ahmad Lone

Abstract:

The Mukherjee-Islam Model is commonly used as a simple life time distribution to assess system reliability. The model exhibits a better fit for failure information and provides more appropriate information about hazard rate and other reliability measures as shown by various authors. It is possible to introduce a location parameter at a time (i.e., a time before which failure cannot occur) which makes it a more useful failure distribution than the existing ones. Even after shifting the location of the distribution, it represents a decreasing, constant and increasing failure rate. It has been shown to represent the appropriate lower tail of the distribution of random variables having fixed lower bound. This study presents the reliability computations and probability weighted moment estimation of three parameter model. A comparative analysis is carried out between three parameters finite range model and some existing bathtub shaped curve fitting models. Since probability weighted moment method is used, the results obtained can also be applied on small sample cases. Maximum likelihood estimation method is also applied in this study.

Keywords: comparative analysis, maximum likelihood estimation, Mukherjee-Islam failure model, probability weighted moment estimation, reliability

Procedia PDF Downloads 241