Suchergebnisse

Cusum charts are widely used for detecting deviations of a process about a target value and also for finding evidence of change in the mean of a process. The testing theory approximates the process by a Wiener process or a Brownian bridge, respectively. For quality control, it is important that other aspects are monitored in addition to or instead of the mean. Here, we show that cusum theory is easily adapted when the target is not the mean but some other aspect of the distribution.

Different change point models for AR(1) processes are reviewed. For some models, the change is in the distribution conditional on earlier observations. For others, the change is in the unconditional distribution. Some models include an observation before the first possible change time – others not. Earlier and new CUSUM type methods are given, and minimax optimality is examined. For the conditional model with an observation before the possible change, there are sharp results of optimality in the literature. The unconditional model with possible change at (or before) the first observation is of interest for applications. We examined this case and derived new variants of four earlier suggestions. By numerical methods and Monte Carlo simulations, it was demonstrated that the new variants dominate the original ones. However, none of the methods is uniformly minimax optimal.

A proposta deste artigo é analisar a aplicação de um gráfico de somas acumuladas (CUSUM) binomial no monitoramento da presença de coliformes totais, um indicador biológico para verificar a qualidade da água de abastecimento em um processo de tratamento de água. A série analisada é proveniente de uma estação de tratamento de água e corresponde ao período de 2007 a 2009, com frequência mensal. O tratamento estatístico dos dados foi realizado no GNU R, com desenvolvimento de rotinas para aproximar o limite superior do gráfico CUSUM binomial. Além disso, foi realizado um estudo comparativo para investigar se existe diferença significativa quanto à sensibilidade existente entre a aplicação do referido gráfico e o tradicional gráfico do tipo Shewhart, geralmente aplicado em monitoramento de processos. Os resultados demonstraram que este estudo foi fundamental para definir a melhor opção entre os gráficos para a análise estatística deste processo.

In recent decades the temperature trends are being confirmed earlier. To determine the temperature changes in Brazil, analysis of annual trends and seasons was carried out based on daily data from 187 weather stations, covering the period from 2000 to 2014. Trends were confirmed using the Cumulative Sum Chart, a method that notices the changes in series faster than others. The modeling of temperature was obtained by Multiple Regression and using Cluster Analysis, and through this it was possible to group weather stations. The observed trends confirmed oscillations in cooling, heating and, in some cases, cooling followed by heating. These main trends presented during the study period were from -15 to 0 in latitude. The region with the highest confirmation of inversions in temperature was the North, then Northeast. Heights below 500 meters presented higher trends. Regarding the seasons, autumn contributed significantly to the trends.

International audience ; The forest decline in tropical areas is one of the largest global environmental threats as the growth of both global population and its needs have put an increasing pressure on these ecosystems. Efforts are ongoing to reduce tropical deforestation rates. Earth observations are increasingly used to monitor deforestation over the whole equatorial area. Change detection methods are mainly applied to satellite optical images which face limitations in humid tropical areas. For instance, due to frequent cloud cover in the tropics, there are often long delays in the detection of deforestation events. Recently, detection methods applied to Synthetic Aperture Radar (SAR) have been developed to address the limitations related to cloud cover. In this study, we present an application of a recently developed change detection method for monitoring forest cover loss from SAR time-series data in tropical zone. The method is based on the Cumulative Sum algorithm (CuSum) combined with a bootstrap analysis. The method was applied to time-series of Sentinel-1 ground range detected (GRD) dual polarization (VV, VH) images forming a dataset of 60 images to monitor forest cover loss in a legal forest concession of the Democratic Republic of Congo during the 2018-2020 period. A cross-threshold recombination was then conducted on the computed maps. Evaluated against reference forest cut maps, an overall accuracy up to 91% and a precision up to 75% in forest clear cut detection was obtained. Our results show that more than 60% of forest disturbances were detected before the PlanetScope-based estimated date of cut, which may suggest the capacity of our method to detect forest degradation.