Research Spotlight

https://www.stat.osu.edu/sites/default/files/chris-hans-banner940px.jpg https://www.stat.osu.edu/sites/default/files/tao-jared-940px_0.jpg

 

Sea Surface Temperatures Anomalies

Dr. Mark Berliner and Dr. Noel Cressie

Shown are sea surface temperatures (SST) anomalies observed in the tropical Pacific in December 2002, compared to the HiDyn-Model SST forecast based on data up to and including May 2002. Details of the Bayesian HiDyn statistical model that gives 7-month forecasts can be found in L.M. Berliner, C.K. Wikle, and N. Cressie (2000), "Long-lead prediction of Pacific SSTs via Bayesian dynamic modeling," Journal of Climate, 13, 3953- 3968. A web-based tool that gives current forecasts is available at the ENSO page on the SSES Program web-site.

Pictures on the right are directly linked to appropriate research page.

 

  Picture for Sea Surface Temperatures Anomalies in Dec. 2002
Space-Time Modeling of Particulate Matter
 

Predicted PM2.5 and PM10 concentration levels on three consecutive days across the state of Ohio using a bivariate dynamic process convolution model. This modeling approach explicitly takes into account the additive nature of the components of particulate matter. The spatial dependence structure of these latent components depends on the size of the particles (small particles can travel farther than large particles) and the direction and strength of the wind. This work is described in C.A. Calder (2003), "A Bayesian Dynamic Process Convolution Approach to Modeling PM2.5 and PM10 Concentration Levels," Department of Statistics Preprint No. 724, The Ohio State University.

Pictures on the right are directly linked to appropriate research page.

 

  Picture for Space-Time Modeling of Particulate Matter

Spatio-Temporal Analysis of Global Remotely-Sensed CO2

Dr. Noel Cressie

Shown are predicted mid-tropospheric CO2 values on May 8 and 16, 2003, based on measurements made by the Atmopheric InfraRed Sounder (AIRS) on board NASA's Aqua satellite. Details of the spatio-temporal random effects statistical model that produced the predictions and their associated prediction standard errors for a period of 16 days, can be found in M. Katzfuss and N. Cressie (2010), "Spatio-temporal smoothing and EM estimation for massive remote-sensing data sets," Journal of Time Series Analysis, 32, 430-446.

Pictures on the right are directly linked to appropriate research page.

 

  Predicted mid-tropospheric CO2 on May 8 and 16, 2003
Total Column Ozone on 10/5/88
 

 Total Column Ozone (TCO) world map for October 5, 1988, from three different projections. The southern-polar projection (top right) shows the ozone hole over Antarctica that is characteristic of this region during the Southern-Hemisphere spring. The quantity mapped is the predicted TCO (here the posterior expectation of TCO given all the spatially irregular and incomplete data observed on October 5, 1988). Details of the hierarchical multi-resolution statistical model that gives the predicted values can be found in G. Johannesson and N. Cressie (2004), "Variance-covariance modeling and estimation for multi-resolution spatial models," in geoENV 2002 - Geostatistics for Environmental Applications, eds X. Sanchez-Vila, J. Carrera, and J. Gomez-Hernandez, Kluwer, Dordrecht, forthcoming.

Pictures on the right are directly linked to appropriate research page.

 

  Picture for Total Column Ozone on 10/5/88
Influenza Epidemic in Scotland 1989-1990
 

Time series of plots of the predicted relative risk (RR) in the 56 health districts of Scotland during the height of an influenza epidemic (Winter of 1989-1990). The "bubbles" are centered at the district centroids and the area of each bubble is proportional to the predicted RR (here the posterior median of the RR given all influenza hospital admissions, by health district, during the epidemic). Details of the Bayesian spatio-temporal model that gives the predicted values can be found in A. Mugglin, N. Cressie, and I. Gemmell (2002), "Hierarchical statistical modelling of influenza epidemic dynamics in space and time," Statistics in Medicine, 21, 2703-2721.

Pictures on the right are directly linked to appropriate research page.

 

  Picture for Influenza Epidemic in Scotland 1989-1990
Multicategory Support Vector Machine Outputs for Cancer Classification
 

 MSVM decision vectors based on gene expression profiles are plotted for 20 test cases of small round blue cell tumors (SRBCTs) of childhood and five non SRBCT cases. Note that the decision vectors are pretty close to their ideal class representation and they result in correct classification for the 20 test cases.
Details of this research can be found in Y. Lee and C.-K. Lee (2003)"Classification of Multiple Cancer Types by Multicategory Support Vector Machines Using Gene Expression Data",Bioinformatics, vol. 19 (9), 1132-1139.

Pictures on the right are directly linked to appropriate research page.

 

  Picture for Multicategory Support Vector Machine Outputs for Cancer Classification
Design and Analysis of Computer Experiments and its Application to Prosthesis Design
 

 The tibial components of a prosthetic knee replacement include the tibial tray (show with stem) and a fitted insert made from a high molecular-weight polyethylene. The design of prosthetic knees and hips is an interdisciplinary activity that can involve mechanical engineers, material scientists, and orthopedic surgeons, and applied statisticians. Statistical methodology allows researchers to evaluate the mechanical properties of prospective prosthetic designs by interpolating the mechanical properties for a test bed set of designs. The choice of the test bed set of designs and the interpolation process are statistical issues that are studied by statistical researchers involved in the design and analysis of computer experiments.

Pictures on the right are directly linked to appropriate research page.

  Picture for Design and Analysis of Computer Experiments and its Application to Prosthesis Design
Statistical Model of Intergranular Corrosion Growth Kinetics
 

 With techniques which measures the time for the fastest-growing localized corrosion site to penetrate foils of various thicknesses, it is found that the growth kinetics of localized corrosion of AA2024-T3 (a high strength aluminum alloy) exhibits a strong anisotropy. This effect occurs as pits initiated on the surface transferred into intergranular corrosion and thus is the result of the microstructure of this type of alloy. Statistic method is developed to model the relationship between the anisotropy of intergranular corrosion kinetics and the microstructure of this alloy. Details of this research could be found in S. Ruan, D. A. Wolfe, W. Zhang, G. S. Frankel (2004) "Statistical Modeling of Minimum Intergranular Corrosion Path Length in High-Strength Aluminum Alloy", Technometrics , 46 (1) pp. 69-75(7). Extended work could be found in S. Ruan, D. A. Wolfe and G. S. Frankel (2004) "Statistical modeling and computer simulation of intergranular corrosion growth in AA2024-T3 aluminum alloy", Journal of Statistical Planning and Inference, 126(2),553-568.

Pictures on the right are directly linked to appropriate research page.

  Picture of Dr. Wolfe's project.