Most of SpaceTimeViz’s research works were motivated by real applications, which belong to the following categories.

1. Extending models for spatio-temporal prediction

  Massive amounts of data have become easily accessible due to recent open data movements. A large portion of these data involves geographical locations. Spatial statistics was the typical method for handling such data, but it has become infeasible due to the computational complexity of existing methods, making it difficult to manage the ever-increasing volume of data being collected. Our recent research focus has been on finding ways to enhance computational efficiency while maintaining the flexibility of the model. We have been engaged in at least the following related areas.


a. Ding-Chih Lin, Hsin-Cheng Huang and ShengLi Tzeng*. (2023). Some enhancements to DeepKriging. STAT, 12(1): e559, https://doi.org/10.1002/sta4.559.


b. ShengLi Tzeng, Hsin-Cheng Huang*. (2018). Resolution adaptive fixed rank kriging. TECHNOMETRICS, 60, 198-208.


c. ShengLi Tzeng, Hsin-Cheng Huang*. (2015). Non-stationary multivariate spatial covariance estimation via low-rank regularization. STATISTICA SINICA, 25, 151-171.


d. ShengLi Tzeng, Hsin-Cheng Huang, Noel Cressie*. (2005). A fast, optimal spatial-prediction method for massive datasets. JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION, 100, 1343-1357.



2. Statistical problems arising from dense air pollution data

  PM2.5 pollution is harmful to human health, and the air quality monitoring stations managed by the Environmental Protection Agency provide high-quality measurements. However, these stations are expensive to establish and are sparsely distributed, with only 77 monitoring stations in Taiwan. In contrast, the recent Airbox monitoring network consists of thousands of low-cost Internet of Things (IoT) sensing devices, offering advantages in terms of spatial coverage. Airbox data are often complex and of unstable quality, which has motivated us to focus on at least the following aspects.


a. ShengLi Tzeng, Chi-Wei Lai and Hsin-Cheng Huang*. (2023). Spatially adaptive calibrations of AirBox PM2.5 data. BIOMETRICS, 79, 3637-3649.


b. Yu-Jie Huang, Shun-Hong Wang and ShengLi Tzeng*. (2022). Exploring the spatial patterns of the PM2.5 time series – A case study of microsensors in Taichung. JOURNAL OF THE CHINESE STATISTICAL ASSOCIATION, 60, 178-194.


c. Guowen Huang, Ling-Jyh Chen, Wen-Han Hwang, ShengLi Tzeng, and Hsin-Cheng Huang*. (2018). Real-time PM 2.5 mapping and anomaly detection from AirBoxes in Taiwan. ENVIRONMETRICS, 29, e2537.



3. Classification/clustering with curve data and images

  Spatial statistics originally dealt with curve or surface interpolation. In biomedical research, it is common to encounter data with curve or surface characteristics, making it possible to apply spatial statistics to specific problems. Our endeavors bridge the gaps between one-dimensional and higher-dimensional functional data analysis, inspiring novel approaches in the field.


a. ShengLi Tzeng, Chun-Shu Chen, Yu-Fen Li and Jin-Hua Chen*. (2023). On summary ROC curve for dichotomous diagnostic studies: an application to meta-analysis of COVID-19. JOURNAL OF APPLIED STATISTICS, 50, 1418-1434.


b. Sean R. Anderson*, Rachael Jocewicz, Alan Kan, Jun Zhu, ShengLi Tzeng and Ruth Y. Litovsky. (2022). Sound source localization patterns and bilateral cochlear implants: Age at onset of deafness effects. PLOS ONE, 17, e0263516.


c. ShengLi Tzeng, Jun Zhu*, Weisman, Amy J. Weisman, Tyler J. Bradshaw, and Robert Jeraj. (2021). Spatial process decomposition for quantitative imaging biomarkers using multiple images of varying shapes. STATISTICS IN MEDICINE, 40, 1243-1261.


d. ShengLi Tzeng, Christian Hennig, Yu-Fen Li, Chien-Ju Lin* (2018). Dissimilarity for functional data clustering based on smoothing parameter commutation. STATISTICAL METHODS IN MEDICAL RESEARCH, 27, 3492-3504.



4. High-dimensional data visualization

   Amid evolving data collection techniques, numerous studies confront the challenge of high-dimensional data due to an abundance of variables surpassing sample sizes. Collaboration on matrix visualization at Academia Sinica has yielded publications illustrating subject-variable relationships with minimal assumptions. The exploration of the hidden structure of time series and spatial data is also a natural challenge faced by SpaceTimeViz.


a.  ShengLi Tzeng, Bo-Yu Chen and Hsin-Cheng Huang*. (2024). Assessing spatial stationarity and segmenting spatial processes into stationary components. JOURNAL OF AGRICULTURAL BIOLOGICAL AND ENVIRONMENTAL STATISTICS, 29, 301-319, 2024.


b. Heng-Hui Lue and ShengLi Tzeng*. (2023). Interpretable, predictive spatio-temporal models via enhanced pairwise directions estimation. JOURNAL OF APPLIED STATISTICS, 50, 2914-2933.


c. ShengLi Tzeng, Han-Ming Wu, Chun-Houh Chen (2009). Selection of proximity measures for matrix visualization of binary data. In Biomedical Engineering and Informatics, 2009. BMEI'09. 2nd International Conference on IEEE, pp. 1-9.


d. Han-Ming Wu, ShengLi Tzeng, Chun-Houh Chen (2008). Matrix visualization. In Handbook of Computational Statistics: Data Visualization, Chen CH, Härdle W, Unwin A (eds). Berlin: Springer-Verlag, pp. 681-708.