SOLICITATION NOTICE
R -- Update Software and Graphical User Interface (GUI) for Receiver Operating Characteristic Analysis (ROC) and Related Analyses Relevant to the Fields of Alcohol, Drug and Psychiatric Epidemiology
- Notice Date
- 3/8/2007
- Notice Type
- Solicitation Notice
- NAICS
- 541511
— Custom Computer Programming Services
- Contracting Office
- Department of Health and Human Services, National Institutes of Health, National Institute on Alcohol Abuse and Alcoholism, R&D Contracts Management Branch, 5635 Fishers Lane Room 3016, MSC 9304, Bethesda, MD, 20892-9304, UNITED STATES
- ZIP Code
- 20892-9304
- Solicitation Number
- NIAAA-07-05
- Response Due
- 4/23/2007
- Point of Contact
- Patricia Hanacek, Contract Specialist, Phone 301-594-6226, Fax 301-443-3891, - Patricia Hanacek, Contract Specialist, Phone 301-594-6226, Fax 301-443-3891,
- E-Mail Address
-
hanacekp@mail.nih.gov, hanacekp@mail.nih.gov
- Description
- The National Institute on Alcohol Abuse and Alcoholism (NIAAA) intends to negotiate on a sole-source basis with the University of Chicago to update software and the graphical user interface (GUI) for receiver operating characteristics (ROC) and related analyses relevant to the fields of alcohol, drug and psychiatric epidemiology. Dr. Charles Metz, Professor of Radiology and Medical Physics at the University of Chicago, is chiefly responsible for development of the binomial model for fitting ROC curves and testing differences between ROC curves using maximum likelihood estimation. Based on his work, ROC analysis has become a standard in medical imaging research and has been applied to a variety of problems in domains from psychophysics to epidemiology. Over the years, numerous advances in statistical algorithms have been introduced by Dr. Metz, including the construction of ROC curves from continuous data and/or correlated tests, proper ROC curve fitting for problematic tests, and partial area under the ROC curve for highly sensitive tests. Some of those advances have been incorporated into ROCKIT software, developed by Dr. Metz and his colleagues at the University of Chicago. Dr. Metz is a founder of the application of ROC technology to medical imaging and over the past 25 years his methodologic advances in this area have far surpassed the work of other scientists engaged in this field of inquiry. He has developed new algorithms for ROC curve fitting that are both reliable and robust, and that provide solutions to real-life problems encountered in medical decision making. The purpose of this project is to upgrade and add new statistical algorithms to existing computer software maintained by Dr. Metz and his colleagues at the University of Chicago. No other source has comparable existing software. No other source has developed algorithms for new methodologies in ROC analysis that are relevant at the current time to the Institute's mission, or more generally to alcohol, drug, and psychiatric epidemiology. No other source can implement Dr. Metz's new ROC algorithms since he has years of experience in simulation testing and understands best the nature of the theoretical models underlying the algorithms. No other source has existing software like ROCKIT that provides the foundation for the inclusion of new and related algorithms. No other source has 25 years of experience implementing ROC algorithms. No other source has the expertise and knowledge of ROC analysis necessary to implement new algorithms into existing software, including ROCKIT, which Dr. Metz himself created. Receiver operating characteristic (ROC) analysis was developed in statistical decision theory and later applied to signal detection theory. Although signal detection theory was applied initially to problems in radar, ROC analysis was successfully used in a large variety of studies in psychology and psychophysics by the mid-1960s. Since that time, ROC analysis has increasingly been applied to medical decision making, largely in radiology, but its full potential in epidemiology has yet to be realized. ROC analysis produces curves that graphically represent sensitivity and specificity, calculated at all possible threshold values. An important aspect of ROC analysis is that it provides a common quantitative index describing the area under an ROC curve. This index summarizes the tradeoffs between sensitivity and specificity at each point along the ROC curve. The need for this overall summary measure of test performance in epidemiology cannot be overstated. At present, ROC models have been developed using both parametric and nonparametric methods and can be used with categorical or quantitative data. Methodologies are available that allow investigators to account for the effects of covariates on test performance, and statistical tests have been developed for the comparison of areas under two ROC curves. This project is motivated by the need for updated software and graphical user interface (GUI) for most common ROC analyses relevant to the fields of alcohol, drug, and psychiatric epidemiology. Requirements and Tasks: The tasks are to be conducted in the context of the pre-existing ROCKIT software to accommodate data sets with sample sizes up to 100,000: cleaning and modularization of the ROCKIT computation engine to ensure its stability for a broad range of data sets and to make it usable with essentially any operating system, including SAS and other statistical software; addition of options such as "proper" ROC curve-fitting models and alternative accuracy indices such as partial area; development of a convenient and reliable GUI that allows a wide variety of input-format options and provides ROC plots as well as quantitative output; refinement and expansion of the current User's Guide to include additional tutorial material on ROC analysis and links to relevant publications that can be obtained online; inclusion of nonparametric approaches to generating and testing differences between ROC curves; and inclusion of existing approaches to ROC regression to account for covariates, preferably logit and probit models. Pursuant to 41 U.S.C. 253(c)(1) as set forth in the Federal Acquisition Regulation (FAR) at 6.302-1, the University of Chicago is the only source that is capable of updating ROC software, the ROC GUI upgrading and adding new statistical software since no other source has comparable existing software. To begin anew with a different source would result in substantial duplication of cost to the Government that would not be recovered through competition as well as unacceptable delays. Interested parties may identify their interest and capability to respond to the requirements as described above. A determination not to compete the proposed work based upon responses to this notice is solely within the discretion of the Contracting Officer. This notice of intent is not an announcement of the availability of a Request for Proposal, nor is an RFP available. See Numbered Note 22. NOTE: THIS NOTICE WAS NOT POSTED TO FEDBIZOPPS ON THE DATE INDICATED IN THE NOTICE ITSELF (08-MAR-2007); HOWEVER, IT DID APPEAR IN THE FEDBIZOPPS FTP FEED ON THIS DATE. PLEASE CONTACT fbo.support@gsa.gov REGARDING THIS ISSUE.
- Web Link
-
Link to FedBizOpps document.
(http://www.fbo.gov/spg/HHS/NIH/NIAAA/NIAAA-07-05/listing.html)
- Record
- SN01247148-F 20070310/070308224910 (fbodaily.com)
- Source
-
FedBizOpps Link to This Notice
(may not be valid after Archive Date)
| FSG Index | This Issue's Index | Today's FBO Daily Index Page |