Semiparametric estimation of INAR models using roughness penalization

Article Statistics & Probability

Penalized likelihood methods for modeling count data

Minh Thu Bui et al.

Summary: This paper examines parameter estimation in count data models using penalized likelihood methods. The study focuses on oral reading fluency (ORF) data, which consists of multiple independent count variables with a moderate sample size. Through simulation study, it is found that penalized likelihood methods lead to significant improvements in estimation accuracy compared to unpenalized maximum likelihood.

JOURNAL OF APPLIED STATISTICS (2023)

添加到收藏夹

Article Statistics & Probability

Integer-valued time series model order shrinkage and selection via penalized quasi-likelihood approach

Xinyang Wang et al.

Summary: This paper introduces a new penalized maximum quasi-likelihood (PMQL) estimation method to address the issues of model selection and parameter estimation in integer-valued time series models. The PMQL estimation shows robustness and effectiveness, as demonstrated through theoretical analysis and numerical simulations. The practicality of the PMQL method is further illustrated by analyzing the Westgren's dataset.

METRIKA (2021)

添加到收藏夹

Article Statistics & Probability

Semiparametric integer-valued autoregressive models on DOUBLE-STRUCK CAPITAL Z

Zhengwei Liu et al.

Summary: The article discusses addressing negative values and negative correlations in integer-valued time series data, proposing a semiparametric model that can handle autoregressive coefficients of arbitrary sign by rounding the innovation distribution. The model resolves issues with parameter estimation and confidence intervals by introducing a log-concave innovation approach.

CANADIAN JOURNAL OF STATISTICS-REVUE CANADIENNE DE STATISTIQUE (2021)

添加到收藏夹

Article Business, Finance

Analysis and Forecasting of Risk in Count Processes

Annika Homburg et al.

Summary: Risk measures are commonly used to prepare for adverse events, but using Gaussian time series models for forecasting count data may distort risk assessment. The performance of count time series analysis methods may be affected by estimation uncertainty and discreteness phenomena.

JOURNAL OF RISK AND FINANCIAL MANAGEMENT (2021)

添加到收藏夹

Article Statistics & Probability