Biostatistics
-
Statistical models are constructed for the drug concentration-time profiles in a pharmacokinetic study. The statistical models help on the evaluation of the safety and effectiveness of the new drug under study in Phase I of a clinical trial. Similarly, bioequivalence tests between the test drug and patent drug are developed based on statistical models for the drug concentration-time profiles in a 2×2 crossover design.
-
AIDS related research, bio-indicators and corresponding diseases research, aging study, and brain learning.
-
Mainly cover biomedical statistics, epidemiology statistical analysis, robust likelihood inference related data, and application of generalized linear models. Develop the parametric robust likelihood function inference without any idea of data distribution.
Industrial Statistics
-
Industrial statistics is the application of statistics in precision industries, the most common one of which is product quality control. Evaluating product quality usually requires product life tests to collect related information after using the product, and then a set of quality control standards are established based on the statistically analyzed data to meet the market demand. On the other hand, to efficiently and precisely control product operation in terms of quality control, the statistical experiment design is often required to evaluate preceding tests for the purpose of cost saving or enhancing the precision of highly reliable evaluations. The institute emphasizes experimental design and the Taguchi method in terms of studies on industrial statistics, but has mainly adopted product reliability analysis in recent years. Following technological development, product reliability has become an important basis for the requirements of consumers and manufacturers toward the products of different brands. Reliability means the probability that a product will be available for use within a specific time under certain conditions. Highly reliable products require a longer time to observe the failure information of the processing life test. Furthermore, due to high production costs, too much time and cost would be spent to observe enough data to inspect product reliability. Therefore, the accelerated life test is generally adopted to shorten the test time. Meanwhile, engineers or quality control personnel usually can provide product reliability associated information with an extensive understanding on the product’s process and quality from practical experiences. Currently, the reliability research of the institute is focused on Bayesian Reliability Analysis for accelerating life tests and competitive risks.
-
Statistical process control (SPC) provides tools for maintaining the quality of products in a mass production process. Basic techniques for SPC include control charts (Shewhart chart, Range chart, np-chart), process capability analysis, tolerance limits and the assessment of the average run length. More research oriented techniques include control charts for correlated data (EWMA chart, Copula-based chart), multivariate control charts, and profile monitoring.
-
Degradation Analysis and Optimal Test Planning in Degradation Tests
-
Traditional accelerated life tests often result in highly censored failure-time data or no observed failures for high-reliability products. Alternatively, conducting a degradation test can measure the time-dependent quality characteristics that are highly related to the failure mechanism of products. The resulting degradation data provide more information compared to failure-time data. By modeling the degradation path and defining the first-hitting-time (or first passage time), the time at which the degradation reaches a critical threshold as the failure time of the product, more accurate and precise estimates of the lifetime distribution can be obtained. The primary statistical methods for degradation analysis include mixed-effect models and stochastic processes.
-
On the other hand, determining an optimal test plan within a cost constraint, thereby precisely estimating the lifetime information and improving product reliability, is another important research topic in reliability analysis.
-
Other case studies of degradation analysis have been applied to batteries and cells, metallic layer resistors, semiconductors and microelectronics, electrical insulations and dielectrics, plastics and polymers, metal fatigue, pharmaceuticals, food and drugs, epidemiology and medical research.
Financial Engineering and Statistics
-
Financial Engineering
-
Derivatives pricing
-
Credit risk and systemic risk
-
Portfolio optimization and hedge
-
Stochastic analysis and stochastic control on finance
-
Statistical learning and machine learning on algorithm trading
-
Statistics and Econometrics
-
Time series analysis
-
High dimensional data analysis
-
Meta analysis
Spatial Statistics
-
Spatial statistics is a branch of statistics dedicated to the analysis of data with geographic location information. Its core lies in modeling spatial dependence and spatial heterogeneity among observations. Unlike time series data, spatial data generally lack a natural ordering, making traditional causal modeling inapplicable. Instead, spatial statistical models are often constructed based on Tobler’s First Law of Geography, which posits that observations at nearby locations tend to be more similar than those farther apart.
-
Research in spatial statistics can be broadly categorized into three major areas: geostatistics, which focuses on interpolation and prediction for continuous spatial variables; spatial lattice models, which analyze data defined over discrete areal units or grid cells, commonly used in disease mapping and image analysis; and spatial point processes, which model the randomness and distributional patterns of events occurring in space. These methodologies are widely applied in meteorology, ecology, geology, environmental monitoring, epidemiology, and image analysis.
-
With the rapid advancement of high-resolution spatial data and GIS technologies, spatial statistics has increasingly integrated Bayesian inference, machine learning, and spatial big data analytics, driving both theoretical and applied developments. As a result, it has become one of the most dynamic and promising research areas in modern statistical science.
