Att C - SAG recommendations

Attachment C- SAG Recommendations for Statistical Software Support and Priorities

Regarding Statistical Software Support and Priorities at CDC/ATSDR
Interim Notes from the Statistical Advisory Group
These notes reflect the intermediate results of SAG’s process in developing guidance for statistical
software support at CDC/ATSDR.
In late 2011, the Statistical Advisory Group (SAG) was asked to provide a rationale for making
decisions about statistical software access. We address a request for guidance on the issue of
EpiInfo as a suitable low-cost alternative to SAS, and the larger issue of top tier statistical software
access.
Cost Considerations
Our current understanding of the licensing cost of SAS at CDC is that the software is obtained for
CDC under a fixed total cost, and that reducing the number of licenses will not alter the total cost –
hence, lowering the number of SAS licenses will merely increase the per-license cost. We
recommend that EITPO distribute the cost of SAS to the CIOs according to FTEs rather than SAS
licenses to eliminate the (false) incentive to restrict SAS licenses.
Regarding EpiInfo
EpiInfo is designed to support the rapid production of data in the field and to allow simple, rapid,
basic analysis of this data. From the EpiInfo website (http://wwwn.cdc.gov/epiinfo/):
Physicians, nurses, epidemiologists, and other public health workers lacking a background in
information technology often have a need for simple tools that allow the rapid creation of data
collection instruments and data analysis, visualization, and reporting using epidemiologic methods.
Epi Info™, a suite of lightweight software tools, delivers core ad-hoc epidemiologic functionality
without the complexity or expense of large, enterprise applications.
EpiInfo serves as a powerful set of tools including data management and analytic support, designed
to permit the rapid acquisition of data in public health field applications and to permit rapid, ad hoc
analysis of such data. EpiInfo does not serve as an alternative for statistical software such as SAS or
Stata.
Regarding a Standard Statistical Software Priority at CDC
Regarding SAS
SAS is the standard software for professional statistical analysis at CDC/ATSDR. SAS provides a full
spectrum of support, from data management and manipulation to basic descriptive analysis to
complex inference and modeling.
SAS provides a stable, common basis for statistical practice and project management, minimizing
training issues and maximizing project continuity. The vast majority of statistical work at

Attachment C- SAG Recommendations for Statistical Software Support and Priorities

CDC/ATSDR can be conducted entirely within the SAS platform. Common proficiency in a standard
software package allows the focus to be on methodology, not platform details.
The use of alternative software for statistical analysis by non-statisticians may be driven by issues
of ease or simplicity of use, as well as potential savings, but bear additional costs and consequences:
The support of alternative software will require additional training and support
requirements.
The use of multiple software packages will complicate the conduct of statistical work.
While the early work in a project may be well served by an alternative software package, it is likely
that substantial projects will grow in complexity, requiring the use of a more complex software
package. Using SAS ensures the continuity of project analyses by supporting project work through
all levels of complexity.
On the Use of Other Software
Several other statistical software packages are permitted at CDC – there are two broad types of
statistical software: comprehensive and application-specific. Comprehensive statistical software
packages (e.g., SPSS, Stata, R) support a full suite of statistical routines and functionality, but
emphasize different end user needs. For example, SPSS is commonly used in Social Science, Stata
supports statistical models and routines frequently used in Economics, and R is preferred by the
academic community as it is public domain and allows new statistical methods to be coded and
shared quickly without the onerous testing and documentation requirements of commercial
statistical software. For a number of statistical applications, there are specialized software
packages: SUDAAN (for complex survey work) and StatXact/LogXact (for small sample or exact
applications) – these merit support as needed – sample survey and small sample analyses are
common special problems in public health applications.
SAG Recommendations
SAG supports a standard statistical software suite for CDC/ATSDR, ensuring a common basis for
data management, coding skillsets and project management.
SAG supports SAS as this standard software suite.
SAG supports the allocation of specialized software for specific applications, such as SUDAAN for
survey analysis and StatXact/LogXact for exact/small sample analysis as needed.
SAG does not support the use of EpiInfo as a replacement for SAS.
SAG believes that a change in standard statistical software suite for CDC/ATSDR should be
discussed at the agency level, not division by division.