arXiv subjects

András Zempléni

Publications and source records attributed to András Zempléni.

13 recordsLinked to original sources

arXiv2025

A new adaptive two-layer model for opinion spread in hypergraphs: parameter sensitivity and estimation

When opinion spread is studied, peer pressure is often modeled by interactions of more than two individuals (higher-order interactions). In our work, we introduce a two-layer random hypergraph model, in which hyperedges represent households and workplaces. Within this overlapping, adaptive structure, individuals react if their opinion is in majority in their groups. The process evolves through random steps: individuals can either change their opinion, or quit their workplace and join another one in which their opinion belongs to the majority. Based on computer simulations, our first goal is to describe the effect of the parameters responsible for the probability of changing opinion and quitting workplace on the homophily and speed of polarization. We also analyze the model as a Markov chain, and study the frequency of the absorbing states. Then, we quantitatively compare how different statistical and machine learning methods, in particular, linear regression, xgboost and a convolutional neural network perform for estimating these probabilities, based on partial information from the process, for example, the distribution of opinion configurations within households and workplaces. Among other observations, we conclude that all methods can achieve the best results under appropriate circumstances, and that the amount of information that is necessary to provide good results depends on the strength of the peer pressure effect.

cs.SI↗

arXiv2024

Parameter estimation of epidemic spread in two-layer random graphs by classical and machine learning methods

Our main goal in this paper is to quantitatively compare the performance of classical methods to XGBoost and convolutional neural networks in a parameter estimation problem for epidemic spread. As we use flexible two-layer random graphs as the underlying network, we can also study how much the structure of the graphs in the training set and the test set can differ while to get a reasonably good estimate. In addition, we also examine whether additional information (such as the average degree of infected vertices) can help improving the results, compared to the case when we only know the time series consisting of the number of susceptible and infected individuals. Our simulation results also show which methods are most accurate in the different phases of the epidemic.

cs.SI↗

arXiv2024

Estimating the parameters of epidemic spread on two-layer random graphs: a classical and a neural network approach

In this paper, we study the spread of a classical SIR process on a two-layer random network, where the first layer represents the households, while the second layer models the contacts outside the households by a random scale-free graph. We build a three-parameter graph, called polynomial model, where the new vertices are connected to the existing ones either uniformly, or preferentially, or by forming random triangles. We examine the effect of the graph's properties on the goodness of the estimation of the infection rate $τ$, which is the most important parameter, determining the reproduction rate of the epidemic. In the classical maximum likelihood approach, to estimate $τ$ one needs to approximate the number of SI edges between households, since the graph itself is supposed to be unobservable. Our simulation study reveals that the estimation is poorer at the beginning of the epidemic, for larger preferential attachment parameter of the graph, and for larger $τ$. We present two heuristic improvement algorithms and establish our method to be robust to changes in average clustering of the graph model. We also extend a graph neural network (GNN) approach for estimating contagion dynamics for our two-layered graphs. We find that dense networks offer better training datasets. Moreover, GNN perfomance is measured better using the $l_2$ loss function rather than cross-entropy.

q-bio.PE↗

arXiv2021

A Note on the Moments of Special Mixture Distributions, with Applications for Control Charts

Control charts can be applied in a wide range of areas, this paper focuses on generalisations suitable for healthcare applications. We concentrate on the effect of using mixture distributions as the possible shifts in the process mean value. We give a closed formula for the expected squared deviation from the target value between samplings. This observation can be utilised in finding the cost-optimal parameters of the control chart.

stat.AP↗

arXiv2020

Trend detection in GEV models

In recent environmental studies extreme events have a great impact. The yearly and monthly maxima of environment related indices can be analysed by the tools of extreme value theory. For instance, the monthly maxima of the fire weather index in British Columbian forests might be modelled by GEV distribution, but the stationarity of the time series is questionable. This property can lead us to different approaches to test if there is a significant trend in past few years data or not. An approach is a likelihood ratio based procedure which has favourable asymptotic properties, but for realistic sample sizes it might have a large error. In this paper we analyse the properties of the likelihood ratio test for extremes by bootstrap simulations and aim to determine a minimal required sample size. With the theoretical results we re-asses the trends of fire weather index in British Columbian forests.

stat.ME↗

arXiv2020

how well can sensitivity and sojourn time be estimated

Chronic disease progression models are governed by three main parameters: sensitivity, preclinical intensity, and sojourn time. The estimation of these parameters helps in optimizing screening programs and examine the improvement in survival. Multiple approaches exist to estimate those parameters. However, these models are based on strong underlying assumptions. The main aim of this article is to investigate the effect of these assumptions. For this purpose, we developed a simulator to mimic a breast cancer screening program directly observing the exact onset and the sojourn time of the disease. We investigate the effects of assuming the sensitivity to be constant, inter-screening interval and misspecifying the sojourn time. Our results indicate a strong correlation between the estimated parameters, and that the chosen sojourn time-distribution has a strong effect on the accuracy of the estimation. These findings shed a light on the seemingly discrepant results got by different authors using the same data sets but different assumptions.

stat.AP↗

arXiv2019

Markov Chain-based Cost-Optimal Control Charts for Healthcare Data

Control charts have traditionally been used in industrial statistics, but are constantly seeing new areas of application, especially in the age of Industry 4.0. This paper introduces a new method, which is suitable for applications in the healthcare sector, especially for monitoring a health-characteristic of a patient. We adapt a Markov chain-based approach and develop a method in which not only the shift size (i.e. the degradation of the patient's health) can be random, but the effect of the repair (i.e. treatment) and time between samplings (i.e. visits) too. This means that we do not use many often-present assumptions which are usually not applicable for medical treatments. The average cost of the protocol, which is determined by the time between samplings and the control limit, can be estimated using the stationary distribution of the Markov chain. Furthermore, we incorporate the standard deviation of the cost into the optimisation procedure, which is often very important from a process control viewpoint. The sensitivity of the optimal parameters and the resulting average cost and cost standard deviation on different parameter values is investigated. We demonstrate the usefulness of the approach for real-life data of patients treated in Hungary: namely the monitoring of cholesterol level of patients with cardiovascular event risk. The results showed that the optimal parameters from our approach can be somewhat different from the original medical parameters.

stat.AP↗

arXiv2018

Mathematical modelling European temperature data: spatial differences in global warming

This paper shows an analysis of the gridded European precipitation data. We combine simple linear regression with data mining tools like clustering, and evaluate the strength of the results by the modern bootstrap methods. We have used the 0.5 grade-grid of daily temperatures for 65 years, created by the European Climate Assessment. We have checked the stability of the results by changing the starting point of the linear regression - this approach might be valuable for climatologists in finding the "best" starting point for assessing the global warming in Europe. Different bootstrap approaches were compared and it turned out that the dependent weighted bootstrap is the best for checking the significance of the estimators.

stat.AP↗

arXiv2018

Multivariate stable distributions and their applications for modelling cryptocurrency-returns

In this paper we extend the known methodology for fitting stable distributions to the multivariate case and apply the suggested method to the modelling of daily cryptocurrency-return data. The investigated time period is cut into 10 non-overlapping sections, thus the changes can also be observed. We apply bootstrap tests for checking the models and compare our approach to the more traditional extreme-value and copula models.

stat.AP↗

arXiv2018

Regression estimator for the tail index

Estimating the tail index parameter is one of the primal objectives in extreme value theory. For heavy-tailed distributions the Hill estimator is the most popular way to estimate the tail index parameter. Improving the Hill estimator was aimed by recent works with different methods, for example by using bootstrap, or Kolmogorov-Smirnov metric. These methods are asymptotically consistent, but for tail index $ξ>1$ and smaller sample sizes the estimation fails to approach the theoretical value for realistic sample sizes. In this paper, we introduce a new empirical method, which can estimate high tail index parameters well and might also be useful for relatively small sample sizes.

stat.ME↗

arXiv2016

Generalised block bootstrap and its use in meteorology

In an earlier paper Rakonczai et al. (2014), we have emphasized the effective sample size for autocorrelated data. The simulations were based on the block bootstrap methodology. However, the discreteness of the usual block size did not allow for exact calculations. In this paper we propose a generalisation of the block bootstrap methodology, relate it to the existing optimisation procedures and apply it to a temperature data set. Our other focus is on statistical tests, where quite often the actual sample size plays an important role, even in case of relatively large samples. This is especially the case for copulas. These are used for investigating the dependencies among data sets. As in quite a few real applications the time dependence cannot be neglected, we investigated the effect of this phenomenon to the used test statistic. The critical values can be computed by the proposed new block bootstrap simulation, where the block sizes are determined e.g. by fitting a VAR model to the observations. The results are illustrated for models of the used temperature data.

math.ST↗

arXiv2013

Applications of threshold models and the weighted bootstrap for Hungarian precipitation data

This paper presents applications of the peaks-over threshold methodology for both the univariate and the recently introduced bivariate case, combined with a novel bootstrap approach. We compare the proposed bootstrap methods to the more traditional profile likelihood. We have investigated 63 years of the European Climate Assessment daily precipitation data for five Hungarian grid points, first separately for the summer and winter months, then aiming at the detection of possible changes by investigating 20 years moving windows. We show that significant changes can be observed both in the univariate and the bivariate cases, the most recent period being the most dangerous, as the return levels here are the highest. We illustrate these effects by bivariate coverage regions.

math.ST↗

arXiv2012

Weighted bootstrap in GARCH models

GARCH models are useful tools in the investigation of phenomena, where volatility changes are prominent features, like most financial data. The parameter estimation via quasi maximum likelihood (QMLE) and its properties are by now well understood. However, there is a gap between practical applications and the theory, as in reality there are usually not enough observations for the limit results to be valid approximations. We try to fill this gap by this paper, where the properties of a recent bootstrap methodology in the context of GARCH modeling are revealed. The results are promising as it turns out that this remarkably simple method has essentially the same limit distribution, as the original estimatorwith the advantage of easy confidence interval construction, as it is demonstrated in the paper. The finite-sample properties of the suggested estimators are investigated through a simulation study, which ensures that the results are practically applicable for sample sizes as low as a thousand. On the other hand, the results are not 100% accurate until sample size reaches 100 thousands - but it is shown that this property is not a feature of our bootstrap procedure only, as it is shared by the original QMLE, too.

math.ST↗