Vision Research Center


News from the VRC

VRC Administration

VRC Facilities

Related Penn Sites


Vision Science Seminar


UPenn - Vision Research Center Biostatistics Technical Description

Core biostatisicians are available to help analyze data, design pilot studies (such as secondary data analysis of publicly available data from clinical trails or observational studies), and develop grant proposal for funding to conduct larger studies. 

The core has extensive experience handling designs that are typical of vision research.  For example, in ophthalmologic and vision research, data are often collected from both eyes of a subject (a patient or an animal): ERG is often measured in both eyes of a dog; visual acuity and OCT are often measured in both eyes of a patient with age-related macular degeneration.

Because data from two eyes of a subject are correlated, such data should not be analyzed in the same way as data from two independent subjects. For example, a standard two group ttest is not valid for comparing means of data from two eyes. However, ophthalmologic investigators and vision scientists may not recognize the need to account for inter-eye correlation or don’t know the methods for doing so. Correlated data are often analyzed incorrectly by either totally ignoring the inter-eye correlation or inefficiently by selecting data from only one eye per subject for analysis. Our biostatisticians have a lot of experience in analyzing the correlated eye data, and would be happy to provide you the statistical help to set up the data in the appropriate format, and select the most appropriate statistical procedure for analyzing correlated eye data. We have provided such statistical consulting services to many patient-oriented researchers or bench scientists.  Guidelines are available for preparing data for the core to analyze. We are happy, however, to discuss alternative data formats, particularly in advance of data collection.

Our biostatisticians can help with the actual data management and statistical analysis of data from either clinical researches or laboratory studies. However, some investigators may like to learn from us on how to do the data management and perform standard statistical analysis by themselves using their preferred statistical software. Our biostatisticians would be happy to provide tutoring through one or more in-person meetings. Our biostatisticians mostly use SAS for statistical analysis. We are also capable of using other statistical software because the statistical principles behind each statistical software are basically the same. We can teach you how to set up the data (such as arrange the data from two eye per record into one eye per record for eye-level analysis), select the appropriate statistical procedure (such as analysis of variance, post-hoc multiple comparisons, generalized estimating equations for adjusting inter-eye correlations, etc), and how to interpret the output from statistical software. If necessary, our biostatistician can even develop some very general statistical programming codes for the standard statistical analysis, based on the specific data type and data structure of your study. You can easily modify the statistical code for analyzing similar data from other studies.  We have provided such statistical mentoring and services on several projects, and some investigators now can do most of the statistical analysis work independently.

In addition, the Department of Ophthalmology has acquired access to the OptumInsight administrative medical claims database, jointly with the Penn Center for Clinical Epidemiology and Biostatistics and the Penn Leonard Davis Institute for Health Economics, and thanks to Joan O’Brien access to this database is available to all Penn vision researchers. The database contains comprehensive information on the healthcare services and expenditures for more than 60 million people covered by the United Health group, a large national US insurer. Included in the dataset is de-identified data on all outpatient visits, procedures performed and their associated diagnoses, dates and diagnoses for every hospital admission and discharge, every filled prescription, all laboratory tests ordered and, in most instances, their associated values. Demographic information and the socio-economic status variables, education and yearly income, plus co-pay and standardized costs for all care given during the years 2000-2012 are also included (updates will be available for later years). Access to this database allows investigation of surgical and drug treatment patterns and comparative effectiveness studies for all ocular conditions and diseases, their associated costs, complication rates after ophthalmic procedures, and risk factors for diseases and complications.

For those of you who heard Brian VanderBeek’s talk at our Fall Vision Dinner, you have some sense of the exciting research that can be done with this database. Our Biostatistics module now includes bioinformatics support for expert programming access to the database. If you have a potential application, please contact Brian VanderBeek ( to discuss whether the database is likely to support your application. If so, the contact Gui-shuang Ying ( to get in touch with the database bioinformatician.