| Source |
| The National Science Foundation (NSF) NSF Survey of Science and Engineering Research Facilities which is conducted biennially by the NSF Division of Science Resources Statistics (SRS). |
| Contact |
|
Leslie Christovich
Director, Academic Infrastructure Project
Research and Development Statistics Program
Division of Science Resources Statistics
National Science Foundation
4201 Wilson Boulevard, Suite 965
Arlington, VA 22230
Telephone: (703) 292-7782
Fax: (703) 292-9091
Internet: lchristo@nsf.gov
|
| URL |
| Source data and documentation are available from the National Science Foundation, Division of Science Resources Statistics. |
| Description |
| These data are collected biennially through the National Science Foundation's (NSF's) congressionally mandated Survey of Science and Engineering Research Facilities (Facilities Survey). The survey begam in 1986 in response to Congress' concern about the state of research facilities at the Nation's colleges and universities. NSF's 1984 reauthorization legislation, P.L. 99-159, mandated a data collection to identify and assess the research facilities needs of academic institutions.
The National Institutes of Health (NIH) have cosponsored all cycles of the survey.
Recognizing the expanding use of networking and computing capacity in conducting research, a new set of questions focusing on computing and networking capacity was added to the FY 2003 Facilities survey.
The 2003 survey was mailed to academic and biomedical institutions in November of 2003 and data collection ended May 21, 2004.
Since these data were last collected in 2001, several changes have been made to the population, some of the survey questions, and the release of public data. Some of the changes include:
Research space is broken out by categories of space: laboratory, laboratory support, offices, and other research space.
The threshold for reporting repair and renovation and construction projects was raised from $100,000 to $250,000. Information on construction projects is now provided for each individual project rather than as a calculation of total costs for all construction projects.
Separate questions ask about all medical school data, replacing questions about biological sciences and medical sciences inside or outside a medical school.
A new section on networking and computing capacity has been added to the survey.
Individual institutional data for most items is being made publicly available. Three survey items will remain confidential: information on indirect costs, animal research space, and condition of facilities. These confidential items are being released only as aggregate totals without individual, institutional breakdowns.
Research-performing institutions are defined as institutions having at least $1 million in R&D expenditures or as having received $1 million in NIH funding. |
| Availability |
| Data are available for the Fiscal Year (FY) 2003 survey. |
| These data are not available by institution or state in WebCASPAR. To obtain data by institution or state, use the NSF Survey of Science and Engineering Research Facilities (Not Weighted or Imputed) data source. |
| Current As Of |
| Data are current as of November 2005. |
| Population Size and Structure |
| The 2003 population consisted of 465 research-performing academic institutions and 191 nonprofit biomedical research institutions in the U.S.
Of the 465 academic institutions, 92 percent returned surveys. Of the 191 biomedical organizations, 94 percent returned surveys.
Research-performing academic institutions were defined as colleges and universities with $1 million or more in research and development (R&D) expenditures. In addition, Historically Black Colleges and Universities (HBCUs) with any R&D expenditures were included in the survey. Each academic institution's level of R&D expenditures was determined by the 2002 NSF Survey of Research and Development Expenditures at Universities and Colleges. Military institutions, Veteran's Administration institutions, and Federally Funded R&D Centers (FFRDCs) were excluded. The biomedical institution frame was a list of nonprofit biomedical research organizations and hospitals in the U.S. that received at least $1 million in NIH research funding in FY 2002.
Johns Hopkins University and Applied Physics Lab completed separate survey forms, but their data were combined on the data file and are treated as a single institution in all published tables and study reports. The final population of 465 counts Johns Hopkins University and Applied Physics Lab as a single institution. |
| Weighting |
| The 2003 Facilities survey attempted to obtain responses from all institutions in the defined population. Consequently, one of the usual sources of survey error, sampling error, is not of concern in this survey. However, as is the case in almost all surveys, nonresponse error is of concern. In the 2003 Facilities survey, 92 percent of all eligible institutions responded.
Weights were used to account for unit nonresponse. The weights for the academic institutions were adjusted for the known number of academic institutions by: expenditure categories (the quintiles of the distribution), census region, control (public/private), whether the institution was a historically black college or university, and whether the institution granted Ph.D. degrees. For the biomedical institutions the only auxiliary variables were the grant amount (quintiles of the distribution) and census region. The minimum weights for both academic and biomedical institutions were constrained to be at least 1.0.
The part 1 data are weighted according to the previously described procedures.
The part 2 data are not weighted due to potential measurement error within the survey responses. It is believed that substantially greater measurement error may exist in the part 2 data because FY 2003 was the first year of implementation of these questions and because of the rapidly changing nature and variability of the part 2 data. |
| Estimation/Imputation |
| Item nonresponse was not imputed for part 2 data.
For part 1 data, a series of logistic regression models and linear regression models were developed and used to impute the values for all missing data for institutions that responded to the survey. The predicted values from these models were used to impute for the missing responses, although in some cases stochastic imputations were used to better reproduce expected distributions. The imputation was done for academic data and biomedical data separately. The models for imputing the academic data were developed first and then similar models were then applied to impute the biomedical data, to the extent possible.
A set of core predictors was used for imputing most items across the two types of institutions, but differences in the available data by type of institution limited this to some degree. For academic institutions, the core predictors were: control (public/private), highest degree granted (doctorate/non doctorate), existence of a medical school, FY 2002 total research and development expenditures (overall), and total NASF. For biomedical institutions, the core predictors were: status as a hospital or other biomedical institution, FY 2002 eligible NIH grant awards, and total NASF.
The items were first classified into two categories based on the item nonresponse rates as those with item nonresponse rate greater than 5 percent and with more than 10 units (institutions) missing, and all other items. For the items with rates of less than 5 percent, the core predictors and other variables needed to preserve any skip patterns were used in the regressions. For the items with higher nonresponse rates and a few key items used for most analyses, exploratory analysis was done to try to improve the model fit for these items by including other predictor variables.
Imputed data are not available in WebCASPAR by institution or state.
The data for Johns Hopkins University includes data for the Applied Physics Laboratory. |