Avaliação de métodos de estimação da variância em amostras complexas

A necessidade de conhecer uma população impulsiona um processo de recolha e análise de informação. Usualmente é muito difícil ou impossível estudar a totalidade da população, daí a importância do estudo com recurso a amostras. Conceber um estudo por amostragem é um processo complexo, desde antes da recolha dos dados até a fase de análise dos mesmos. Na maior parte dos estudos utilizam-se combinações de vários métodos probabilísticos de amostragem para seleção de uma amostra, que se pretende representativa da população, denominado delineamento de amostragem complexo. O conhecimento dos erros de amostragem é necessário à correta interpretação dos resultados de inquéritos e à avaliação dos seus planos de amostragem. Em amostras complexas, têm sido usadas aproximações ajustadas à natureza complexa do plano da amostra para a estimação da variância, sendo as mais utilizadas: o método de linearização Taylor e as técnicas de reamostragem e replicação. O principal objetivo deste trabalho é avaliar o desempenho dos estimadores usuais da variância em amostras complexas. Inspirado num conjunto de dados reais foram geradas três populações com características distintas, das quais foram sorteadas amostras com diferentes delineamentos de amostragem, na expectativa de obter alguma indicação sobre em que situações se deve optar por cada um dos estimadores da variância. Com base nos resultados obtidos, podemos concluir que o desempenho dos estimadores da variância da média amostral de Taylor, Jacknife e Bootstrap varia com o tipo de delineamento e população. De um modo geral, o estimador de Bootstrap é o menos preciso e em delineamentos estratificados os estimadores de Taylor e Jackknife fornecem os mesmos resultados; Evaluation of variance estimation methods in complex samples ABSTRACT: The need to know a population drives a process of collecting and analyzing information. Usually is to hard or even impossible to study the whole population, hence the importance of sampling. Framing a study by sampling is a complex process, from before the data collection until the data analysis. Many studies have used combinations of various probabilistic sampling methods for selecting a representative sample of the population, calling it complex sampling design. Knowledge of sampling errors is essential for correct interpretation of the survey results and evaluation of the sampling plans. In complex samples to estimate the variance has been approaches adjusted to the complex nature of the sample plane. The most common are: the linearization method of Taylor and techniques of resampling and replication. The main objective of this study is to evaluate the performance of usual estimators of the variance in complex samples. Inspired on real data we will generate three populations with distinct characteristics. From this populations will be drawn samples using different sampling designs. In the end we intend to get some lights about in which situations we should opt for each one of the variance estimators. Our results show that the performance of the variance estimators of sample mean Taylor, Jacknife and Bootstrap varies with the design and population. In general, the Bootstrap estimator is less precise and in stratified design Taylor and Jackknife estimators provide the same results.

Customer Loyalty toward grocery retailing service: a multidimensional approach to costumer perceived value

Structured abstract Purpose: To deepen, in grocery retail context, the roles of consumer perceived value and consumer satisfaction, as antecedents’ dimensions of customer loyalty intentions. Design/Methodology/approach: Also employing a short version (12-items) of the original 19-item PERVAL scale of Sweeney & Soutar (2001), a structural equation modeling approach was applied to investigate statistical properties of the indirect influence on loyalty of a reflective second order customer perceived value model. The performance of three alternative estimation methods was compared through bootstrapping techniques. Findings: Results provided i) support for the use of the short form of the PERVAL scale in measuring consumer perceived value; ii) the influence of the four highly correlated independent latent predictors on satisfaction was well summarized by a higher-order reflective specification of consumer perceived value; iii) emotional and functional dimensions were determinants for the relationship with the retailer; iv) parameter’s bias with the three methods of estimation was only significant for bootstrap small sample sizes. Research limitations:/implications: Future research is needed to explore the use of the short form of the PERVAL scale in more homogeneous groups of consumers. Originality/value: Firstly, to indirectly explain customer loyalty mediated by customer satisfaction it was adopted a recent short form of PERVAL scale and a second order reflective conceptualization of value. Secondly, three alternative estimation methods were used and compared through bootstrapping and simulation procedures.

Prediction Nomogram for 68Ga-PSMA-11 PET/CT in different clinical settings of PSA failure after radical treatment for Prostate Cancer

Objective The objective of this study was to develop a clinical nomogram to predict gallium-68 prostate-specific membrane antigen positron emission tomography/computed tomography (68Ga-PSMA-11-PET/CT) positivity in different clinical settings of PSA failure. Materials and methods Seven hundred three (n = 703) prostate cancer (PCa) patients with confirmed PSA failure after radical therapy were enrolled. Patients were stratified according to different clinical settings (first-time biochemical recurrence [BCR]: group 1; BCR after salvage therapy: group 2; biochemical persistence after radical prostatectomy [BCP]: group 3; advanced stage PCa before second-line systemic therapies: group 4). First, we assessed 68Ga-PSMA-11-PET/CT positivity rate. Second, multivariable logistic regression analyses were used to determine predictors of positive scan. Third, regression-based coefficients were used to develop a nomogram predicting positive 68Ga-PSMA-11-PET/CT result and 200 bootstrap resamples were used for internal validation. Fourth, receiver operating characteristic (ROC) analysis was used to identify the most informative nomogram’s derived cut-off. Decision curve analysis (DCA) was implemented to quantify nomogram’s clinical benefit. Results 68Ga-PSMA-11-PET/CT overall positivity rate was 51.2%, while it was 40.3% in group 1, 54% in group 2, 60.5% in group 3, and 86.9% in group 4 (p < 0.001). At multivariable analyses, ISUP grade, PSA, PSA doubling time, and clinical setting were independent predictors of a positive scan (all p ≤ 0.04). A nomogram based on covariates included in the multivariate model demonstrated a bootstrap-corrected accuracy of 82%. The nomogram-derived best cut-off value was 40%. In DCA, the nomogram revealed clinical net benefit of > 10%. Conclusions This novel nomogram proved its good accuracy in predicting a positive scan, with values ≥ 40% providing the most informative cut-off in counselling patients to 68Ga-PSMA-11-PET/CT. This tool might be important as a guide to clinicians in the best use of PSMA-based PET imaging.

Frequency domain analysis of stationary time series

This thesis provides a necessary and sufficient condition for asymptotic efficiency of a nonparametric estimator of the generalised autocovariance function of a Gaussian stationary random process. The generalised autocovariance function is the inverse Fourier transform of a power transformation of the spectral density, and encompasses the traditional and inverse autocovariance functions. Its nonparametric estimator is based on the inverse discrete Fourier transform of the same power transformation of the pooled periodogram. The general result is then applied to the class of Gaussian stationary ARMA processes and its implications are discussed. We illustrate that for a class of contrast functionals and spectral densities, the minimum contrast estimator of the spectral density satisfies a Yule-Walker system of equations in the generalised autocovariance estimator. Selection of the pooling parameter, which characterizes the nonparametric estimator of the generalised autocovariance, controlling its resolution, is addressed by using a multiplicative periodogram bootstrap to estimate the finite-sample distribution of the estimator. A multivariate extension of recently introduced spectral models for univariate time series is considered, and an algorithm for the coefficients of a power transformation of matrix polynomials is derived, which allows to obtain the Wold coefficients from the matrix coefficients characterizing the generalised matrix cepstral models. This algorithm also allows the definition of the matrix variance profile, providing important quantities for vector time series analysis. A nonparametric estimator based on a transformation of the smoothed periodogram is proposed for estimation of the matrix variance profile.

Advanced numerical modeling of masonry vaulted structures based on BIM parametric geometry generation

Vaults are an architectural element which during construction history have been built with a great variety of different materials, shapes, and sizes. The shape of these structural elements was often dependent by the necessity to cover complex spaces, by the needed loading capacity, or by architectural aesthetics. Within this complex scenario masonry patterns generates also different effects on loading capacity, load percolation and stiffness of the structure. These effects were been extensively investigated, both with empirical observations and with modern numerical methods. While most of them focus on analyzing the load bearing capacity or the texture effect on vaulted structures, the aim of this analysis is to investigate on the effects of the variation of a single structural characteristic on the load percolation in the vault. Moreover, an additional purpose of the work is related to the coding of a parametrical model aiming at generating different masonry vaulted structures. Nevertheless, proposed script can generate different typology of vaulted structure basing on some structural characteristics, such as the span and the length to cover and the dimensions of the blocks.