DOCSLIB.ORG

1988: Sudaan: a Comprehensive Software Package For

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
1988: Sudaan: a Comprehensive Software Package For SUDAAN: A COMPREHENSIVE SOFTWARE PACKAGE FOR SURVEY DATA ANALYSIS Lisa M. LaVange, Babubhai V. Shah, Beth G. Barnwell and Joyce F. Killinger Lisa M. LaVan~e. Research Triangle Institute KEYWORDS: variance estimation, user interface, The system design for the comprehensive C language survey data analysis software, SUDAAN, evolved with these goals in mind. The first goal is ABSTRACT being met by writing the new software entirely The development of a survey data analysis in the C programming language and testing simul- software package that is flexible enough to meet taneously on several systems. Compilers for the the needs of survey statisticians, as well as C language are available on a variety of operat- research scientists not necessarily schooled in ing systems. The initial programs are currently survey sampling, is underway at the Research executing under the TURBO C compiler on an IBM Triangle Institute. The system was designed to personal computer, the SVC compiler on a 68020, provide significant enhancements to RTI's exist- and under the UNIX operating system on a GOULD ing software package, SUDAAN. The new SUDAAN computer. Versions are also being tested on the system is written in the C language and provides VAX/VMS and IBM/VS. greater efficiency and portability. The soft- The second goal is being met by building ware development strategy consists of developing checks into the system to assure that defini- and debugging stand alone statistical procedures tions of objects are meaningful and computations on an IBM PC with expanded capacity. The proce- are feasible. We have implemented numerically dures are then implemented in two additional stable procedures for the accumulation of sums computing environments, VAX/VMS and IBM/0S. of squares and cross products and for matrix Variance estimation options and data analysis operations, including matrix inversion. techniques not previously available in the RTI For achieving the third goal of improved software package are among the many modifica- efficiency, special care has been taken in the tions incorporated into the new SUDAAN system. design to isolate tasks that are repeated numer- The purpose of this paper is to describe the ous times, such as a numerical calculation on design and development of this system. each variable of each record. These sections of code are optimized to the extent possible with- INTRODUCTION out using assembly language. Some of the opti- Scientists at the Research Triangle Institute mizing algorithms include processing of sparse have been involved in designing, developing, and matrices and vectors, replacement of multiplica- maintaining software systems for the analysis of tion by table lookup with addition, elimination survey data for the past 17 years. Recently, of repetitive (in the loop) "if" tests by RTI has been under contract to the Public Health constructing separate do loops, and many others. Service (PHS) to develop a comprehensive soft- An important product of the SUDAAN design ware package that meets the needs of statistical effort has been the creation of a high level analysts at the National Center for Health statistical programming language. The SUDAAN Statistics (NCHS) and PHS. In particular, this language has been developed to provide the package must be flexible enough to handle most statistician/programmer with easy access to the of the statistical designs used by these agen- software package, thus accomplishing the fourth cies. For this purpose, we have embarked on the goal listed above. Statistical procedures are task of developing a system that incorporates currently being written in the SUDAAN language. many of the features of RTI's existing survey Once they are finalized, some parts may be data analysis system but also includes signifi- converted to C programs to maximize computing cant enhancements. The purpose of this paper is efficiency. to describe in detail the design and development The SUDAAN design allows for four types of of this comprehensive software package. The user interface as depicted in Figure I. The immediate ojective of the SUDAAN system design general user will be able to access the system was to develop a series of procedures capable of through user friendly procedures. Ultimately performing statistical data analysis for complex the procedures will be available through a sample surveys. The ultimate objective was to statistical system, such as SAS, but the current have a system that could aid in the design and design consists of stand-alone procedures that evaluation of new statistical techniques. interface with a standard ASCII file format. RTI's existing software system consists of a The more sophisticated user can write SUDAAN series of procedures that produce statistical programs for computations not available in the analyses in the form of weighted cross-tabula- procedure library. This access feature is tions, generalized ratio estimators, and linear particularly well suited for testing modifica- and logistic regressions. All of the procedures tions to existing procedures, such as the calcu- were constructed using a unified set of FORTRAN lation of an alternative test statistic. A C subroutines that execute in the SAS computing programmer can access the system directly environment on an IBM mainframe. In the course through C callable functions. of developing a new system, we went through the The key features of the SUDAAN system are process of rethinking our current design. Four described in detail in the following sections. goals emerged: RTI SURVEY DATA ANALYSIS PROCEDURES Portability BACKGROUND Reliability/numerical accuracy The collaboration of statisticians and Computational efficiency Ease of modification/enhancement. 761 computer scientists at RTI, beginning in 1971 quantities are used to produce Yates-Grundy-Sen (e.g., Folsom, Bayless, and Shah, 1971) and variance estimates for this design option. The continuing today, has resulted in a series of three variance options cover a wide range of programs for survey data analysis that reflects sample designs including sampling with probabil- state-of-the-art methodology in this field. All ities proportional to size at the first stage. of RTI's analysis packages employ the Taylor Unequal probabilities of selection at a subse- series or Delta method of variance estimation quent stage can be handled by forming pseudo for complex sample survey designs (Folsom, strata at that stage since probabilities of 1974). Each program currently marketed is selection can vary from stratum to stratum. In designed to run within the SAS framework. addition to these variance options, the user may RTI is currently marketing the following request that the variance-covariance matrix for procedures for general use: estimates within a table be computed for all • SESUDAAN procedures allowing for cross-tabulations. • RATIOEST Estimates of variance-covariance matrices were • RTIFREQS previously available only in the SURREGR and • SURREGR RTILOGIT procedures. " RTILOGIT. Phase I software consists of three proce- dures: CROSSTAB, DESCRIPT, and RATIO. The The SESUDAAN (Shah, 1981) procedure produces CROSSTAB procedure is the RTIFREQS equivalent in weighted estimates of means, totals, and propor- the new SUDAAN system. In addition to computing tions and their standard errors. RATIOEST weighted frequency and percentage distributions, (Shah, 1982) calculates weighted frequency dis- Chi-square tests of independence for a two-way tributions and percentages for cross-tabula- contingency table are produced. The tests are tions. computed using a Wald statistic according to RTI's survey regression package, SURREGR methods advocated by Koch, Freeman, and Freeman (Holt, 1977), provides estimates of linear (1976). During Phase II, the Rao-Scott regression coefficients, their estimated Satterthwaite corrected test statistic (Thomas variance-covariance matrix, and tests of hypo- and Rao, 1984 and 1985) will be computed in theses concerning the coefficients that are addition to the Wald statistic for goodness of appropriate for a complex sample design. Logis- fit tests. This statistic has been shown by tic regression capabilities became available Thomas and Rao to have the best characteristics with the development of RTILOGIT (Shah, et al., with respect to both power and significance 1984). This procedure runs in conjunction with level when compared with other statistics SAS PROC LOGIST and produces pseudo maximum correcting for the liberalness of the Wald test. likelihood estimates of the parameters and tests The DESCRIPT procedure computes estimates of of hypotheses about these parameters. means, totals, proportions, geometric means, quantiles, and their sampling errors. Estimates New System Procedures can be requested for the subdomains of user- The SUDAAN system design consists of two specified cross-tabulations. Options are also phases of software development, each resulting available to the user for standardized and/or in a series of statistical procedures. The post-stratified means and proportions and their Phase I procedures are primarily concerned with sampling errors. The standardizing and/or post- descriptive data analysis, including cross-tabu- stratifying variables and their distributions lations, generalized ratio estimation, and are required as input. quantile estimation. The Phase II design, RTI has recently completed a simulation study currently in the planning stages, will incorpor- comparing two methods for estimating the vari- ate several analytical procedures, including
Load more

Recommended publications
  • 1 Computing Variances from Data with Complex Sampling Designs: A
    Computing Variances from Data with Complex Sampling Designs: A Comparison of Stata and SPSS North American Stata Users Group March 12-13, 2001 Alicia C. Dowd, Assistant Professor Univ. Mass. Boston, Graduate College of Education Wheatley Hall, 100 Morrissey Blvd. Boston MA 02125 [email protected] 617 287-7593 phone 617 287-7664 fax Michael B. Duggan, Director of Enrollment Research Suffolk University 8 Ashburton Place Boston MA 02108 [email protected] 617 573-8468 phone 617 720-0970 fax 1 Introduction The National Center for Education Statistics (NCES) is responsible for collecting, analyzing, and reporting data related to education in the United States and other countries (U.S. Department of Education, 1996, p. 2). Among the surveys conducted by the NCES, several pertain to postsecondary schooling and outcomes and are of interest to higher education researchers. These include Beginning Postsecondary Students (BPS), Baccalaureate and Beyond (B&B), National Postsecondary Student Aid Study (NPSAS), the National Study of Postsecondary Faculty (NSOPF), and the Integrated Postsecondary Education Data Set (IPEDS). With the exception of IPEDS, these surveys are conducted using complex survey designs, involving stratification, clustering, and unequal probabilities of case selection. Researchers analyzing these data must take the complex sampling designs into account in order to estimate variances accurately. Novice researchers and doctoral students, particularly those in colleges of education, will likely encounter issues surrounding the use of complex survey data for the first time if they undertake to analyze NCES data. Doctoral programs in education typically have minimal requirements for the study of statistics, and statistical theories are usually learned based on the assumption of a simple random sample.
    [Show full text]
  • Kwame Nkrumah University of Science and Technology, Kumasi
    KWAME NKRUMAH UNIVERSITY OF SCIENCE AND TECHNOLOGY, KUMASI, GHANA Assessing the Social Impacts of Illegal Gold Mining Activities at Dunkwa-On-Offin by Judith Selassie Garr (B.A, Social Science) A Thesis submitted to the Department of Building Technology, College of Art and Built Environment in partial fulfilment of the requirement for a degree of MASTER OF SCIENCE NOVEMBER, 2018 DECLARATION I hereby declare that this work is the result of my own original research and this thesis has neither in whole nor in part been prescribed by another degree elsewhere. References to other people’s work have been duly cited. STUDENT: JUDITH S. GARR (PG1150417) Signature: ........................................................... Date: .................................................................. Certified by SUPERVISOR: PROF. EDWARD BADU Signature: ........................................................... Date: ................................................................... Certified by THE HEAD OF DEPARTMENT: PROF. B. K. BAIDEN Signature: ........................................................... Date: ................................................................... i ABSTRACT Mining activities are undertaken in many parts of the world where mineral deposits are found. In developing nations such as Ghana, the activity is done both legally and illegally, often with very little or no supervision, hence much damage is done to the water bodies where the activities are carried out. This study sought to assess the social impacts of illegal gold mining activities at Dunkwa-On-Offin, the capital town of Upper Denkyira East Municipality in the Central Region of Ghana. The main objectives of the research are to identify factors that trigger illegal mining; to identify social effects of illegal gold mining activities on inhabitants of Dunkwa-on-Offin; and to suggest effective ways in curbing illegal mining activities. Based on the approach to data collection, this study adopts both the quantitative and qualitative approach.
    [Show full text]
  • Handbook on Precision Requirements and Variance Estimation for ESS Households Surveys
    ISSN 1977-0375 Methodologies and Working papers Handbook on precision requirements and variance estimation for ESS households surveys 2013 edition Methodologies and Working papers Handbook on precision requirements and variance estimation for ESS household surveys 2013 edition Europe Direct is a service to help you find answers to your questions about the European Union. Freephone number (*): 00 800 6 7 8 9 10 11 (*)0H The information given is free, as are most calls (though some operators, phone boxes or hotels may charge you). More information on the European Union is available on the Internet (http://europa.eu). Cataloguing data can be found at the end of this publication. Luxembourg: Publications Office of the European Union, 2013 ISBN 978-92-79-31197-0 ISSN 1977-0375 doi:10.2785/13579 Cat. No: KS-RA-13-029-EN-N Theme: General and regional statistics Collection: Methodologies & Working papers © European Union, 2013 Reproduction is authorised provided the source is acknowledged. Acknowledgments Acknowledgments The European Commission expresses its gratitude and appreciation to the following members of the ESS1 Task Force, for their work on Precision Requirements and Variance Estimation for Household Surveys: Experts from the European Statistical System (ESS): Martin Axelson Sweden — Statistics Sweden Loredana Di Consiglio Italy — ISTAT Kari Djerf Finland — Statistics Finland Stefano Falorsi Italy — ISTAT Alexander Kowarik Austria — Statistics Austria Mārtiņš Liberts Latvia — CSB Ioannis Nikolaidis Greece — EL.STAT Experts from European
    [Show full text]
  • Approaches to Imputing Missing Data in Complex Survey Data
    Introduction Offerings Examples Results Conclusions Ending Approaches to imputing missing data in complex survey data Christine Wells, Ph.D. IDRE UCLA Statistical Consulting Group July 27, 2018 Christine Wells, Ph.D. Imputing missing data in complex survey data 1/ 28 Introduction Offerings Examples Results Conclusions Ending Three types of missing data with item non-response Missing completely at random (MCAR) Not related to observed values, unobserved values, or the value of the missing datum itself Missing at random (MAR) Not related to the (unobserved) value of the datum, but related to the value of observed variable(s) Missing not at random (MNAR) The value of the missing datum is the reason it is missing Each variable can have its own type of missing data mechanism; all three can be present in a given dataset Most imputation techniques only appropriate for MCAR and MAR data Christine Wells, Ph.D. Imputing missing data in complex survey data 2/ 28 Introduction Offerings Examples Results Conclusions Ending Different approaches to imputing missing complex survey data Stata: multiple imputation (mi) (and possibly full information maximum likelihood (FIML)) SAS: Four types of hotdeck imputation Fully efficient fractional imputation (FEFI) 2 stage FEFI Fractional hotdeck Hotdeck SUDAAN: Four methods Cox-Iannacchione weighted sequential hotdeck (WSHD) Cell mean imputation Linear regression imputation Logistic regression imputation Christine Wells, Ph.D. Imputing missing data in complex survey data 3/ 28 Introduction Offerings Examples Results Conclusions Ending Handling imputation variation Stata Multiple complete datasets SAS Imputation-adjusted replicate weights (not with hotdeck) BRR (Fay), Jackknife, Bootstrap Multiple imputation (only with hotdeck) SUDAAN Multiple versions of imputed variable (WSHD only) Christine Wells, Ph.D.
    [Show full text]
  • 2015 National Survey on Drug Use and Health
    2015 NATIONAL SURVEY ON DRUG USE AND HEALTH METHODOLOGICAL SUMMARY AND DEFINITIONS DISCLAIMER SAMHSA provides links to other internet sites as a service to its users and is not responsible for the availability or content of these external sites. SAMHSA, its employees, and contractors do not endorse, warrant, or guarantee the products, services, or information described or offered at these other internet sites. Any reference to a commercial product, process, or service is not an endorsement or recommendation by SAMHSA, its employees, or contractors. For documents available from this server, the U.S. Government does not warrant or assume any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed. Substance Abuse and Mental Health Services Administration Center for Behavioral Health Statistics and Quality Rockville, Maryland September 2016 2015 NATIONAL SURVEY ON DRUG USE AND HEALTH: METHODOLOGICAL SUMMARY AND DEFINITIONS Contract No. HHSS283201300001C RTI Project No. 0213984.004.107.006.004 RTI Project Director: David Hunter SAMHSA Project Officer: Peter Tice For questions about this report, please e-mail [email protected]. Prepared for Substance Abuse and Mental Health Services Administration, Rockville, Maryland Prepared by RTI International, Research Triangle Park, North Carolina September 2016 Recommended Citation: Center for Behavioral Health Statistics and Quality. (2016). 2015 National Survey on Drug Use and Health: Methodological summary
    [Show full text]
  • Security Systems Services World Report
    Security Systems Services World Report established in 1974, and a brand since 1981. www.datagroup.org Security Systems Services World Report Database Ref: 56162 This database is updated monthly. Security Systems Services World Report SECURITY SYSTEMS SERVICES WORLD REPORT The Security systems services Report has the following information. The base report has 59 chapters, plus the Excel spreadsheets & Access databases specified. This research provides World Data on Security systems services. The report is available in several Editions and Parts and the contents and cost of each part is shown below. The Client can choose the Edition required; and subsequently any Parts that are required from the After-Sales Service. Contents Description ....................................................................................................................................... 5 REPORT EDITIONS ........................................................................................................................... 6 World Report ....................................................................................................................................... 6 Regional Report ................................................................................................................................... 6 Country Report .................................................................................................................................... 6 Town & Country Report ......................................................................................................................
    [Show full text]
  • Some SAS-SUDAAN Comparisons
    NESUG 17 Analysis Variance Estimation With Complex Surveys: Some SAS-SUDAAN Comparisons Xiuhua Chen and Paul Gorrell Social & Scientific Systems, Inc. Abstract In version 8, SAS® introduced new procedures such as SURVEYMEANS and SURVEYREG for analyzing data from complex surveys (SURVEYFREQ and SURVEYLOGISTIC are part of V9). This paper compares the SAS and SUDAAN® implementation of the Taylor linearization method for variance estimation. Although both SAS and SUDAAN make use of this method (currently it's the only method available in SAS), there are important differences, such as assumptions concerning missing values (nonresponse) and subgroup analysis. Focusing on SURVEYMEANS, we will use concrete examples from two national healthcare databases (the Medical Expenditure Panel Survey and the HCUP Nationwide Inpatient Sample) to clarify these differences. Along the way we'll illustrate the syntax and keywords used in the SAS survey analysis PROCs (especially the STRATA, CLUSTER and DOMAIN statements). This paper should be useful for anyone analyzing data from complex surveys, especially SAS programmers who use SUDAAN for variance estimation. Introduction In our work we use SAS for a wide variety of tasks: restructuring data files, editing variables, statistical analysis, table and report generation, etc. Until recently we did not use SAS for variance estimation with complex survey data because SAS statistical procedures did not have the functionality for incorporating the design properties of complex surveys into the analysis, i.e. they assumed the design to be a simple random sample of an infinite population. In our experience this assumption generally leads to an underestimation of variance. To correctly estimate variance for complex surveys we have generally used SAS-callable SUDAAN whenever we needed to, for example, compute standard errors.
    [Show full text]
  • Download This PDF File
    Control Theory and Informatics www.iiste.org ISSN 2224-5774 (print) ISSN 2225-0492 (online) Vol 2, No.1, 2012 The Trend Analysis of the Level of Fin-Metrics and E-Stat Tools for Research Data Analysis in the Digital Immigrant Age Momoh Saliu Musa Department of Marketing, Federal Polytechnic, P.M.B 13, Auchi, Nigeria. Tel: +2347032739670 e-mail: [email protected] Shaib Ismail Omade (Corresponding author) Department of Statistics, School of Information and Communication Technology, Federal Polytechnic, P.M.B 13, Auchi, Nigeria. Phone: +2347032808765 e-mail: [email protected] Abstract The current trend in the information technology age has taken over every spare of human discipline ranging from communication, business, governance, defense to education world as the survival of these sectors depend largely on the research outputs for innovation and development. This study evaluated the trend of the usage of fin-metrics and e-stat tools application among the researchers in their research outputs- digital data presentation and analysis in the various journal outlets. The data used for the study were sourced primarily from the sample of 1720 out of 3823 empirical on and off line journals from various science and social sciences fields. Statistical analysis was conducted to evaluate the consistency of use of the digital tools in the methodology of the research outputs. Model for measuring the chance of acceptance and originality of the research was established. The Cockhran test and Bartlet Statistic revealed that there were significant relationship among the research from the Polytechnic, University and other institution in Nigeria and beyond. It also showed that most researchers still appeal to manual rather than digital which hampered the input internationally and found to be peculiar among lecturers in the system that have not yet appreciate IT penetration in Learning.
    [Show full text]
  • Strengths and Limitations of Using SUDAAN, Stata, and Wesvarpc for Computing Variances from NCES Data Sets
    NATIONAL CENTER FOR EDUCATION STATISTICS Working Paper Series The Working Paper Series was initiated to promote the sharing of the valuable work experience and knowledge reflected in these preliminary reports. These reports are viewed as works in progress, and have not undergone a rigorous review for consistency with NCES Statistical Standards prior to inclusion in the Working Paper Series. U. S. Department of Education Office of Educational Research and Improvement NATIONAL CENTER FOR EDUCATION STATISTICS Working Paper Series Strengths and Limitations of Using SUDAAN, Stata, and WesVarPC for Computing Variances from NCES Data Sets Working Paper No. 2000-03 February 2000 Contact: Ralph Lee Statistical Standards Program E-mail: [email protected] U. S. Department of Education Office of Educational Research and Improvement U.S. Department of Education Richard W. Riley Secretary Office of Educational Research and Improvement C. Kent McGuire Assistant Secretary National Center for Education Statistics Gary W. Phillips Acting Commissioner The National Center for Education Statistics (NCES) is the primary federal entity for collecting, analyzing, and reporting data related to education in the United States and other nations. It fulfills a congressional mandate to collect, collate, analyze, and report full and complete statistics on the condition of education in the United States; conduct and publish reports and specialized analyses of the meaning and significance of such statistics; assist state and local education agencies in improving their statistical systems; and review and report on education activities in foreign countries. NCES activities are designed to address high priority education data needs; provide consistent, reliable, complete, and accurate indicators of education status and trends; and report timely, useful, and high quality data to the U.S.
    [Show full text]
  • Describing the PRAMS Sample Design for SUDAAN, SAS Complex Survey, SPSS Complex Samples Modules, and STATA
    Describing the PRAMS Sample Design for SUDAAN, SAS Complex Survey, SPSS Complex Samples Modules, and STATA CDC PRAMS has performed the following actions while constructing an analysis dataset for external researchers. External researchers may skip this step and proceed directly to SUDAAN/SAS/SPSS/STATA analysis (step II below). I. Combine the single-year/single-state weighted PRAMS SAS datasets into a single analysis dataset. A. Restrict the dataset to respondents only: IF INQX=1; B. Combine the state stratification scheme (variable STRATUMC) and sample year (variable NEST_YR) into a single variable: SUD_NEST = (STRATUMC*10000) + NEST_YR; C. Sort the dataset by new variable SUD_NEST: PROC SORT DATA=<pramsdata>; BY SUD_NEST; II. Describe the PRAMS Sample Design to the analysis software. A. SUDAAN 1. Use the following statements when using SUDAAN modules: proc crosstab data=<pramsdata> design=strwor; nest sud_nest; totcnt totcnt; samcnt samcnt; weight wtanal; B. SAS Complex Survey 1. Construct a new SAS dataset describing the population count for each sampling stratum. The new dataset will have one observation per sampling stratum and two variables: SUD_NEST (the sampling stratum) and _TOTAL_ (the population count). One method for creating this dataset is as follows: Data totals_for_sas; Set <pramsdata> (keep= sud_nest totcnt); By sud_nest; If first.sud_nest; _total_=totcnt; Keep sud_nest _total_; Run; 2. Use the following statements when using SAS Complex Survey modules: proc surveyfreq data=<pramsdata> totals=totals_for_sas; strata sud_nest; weight wtanal; C. SPSS Complex Samples 1. To describe the PRAMS sample design, open your PRAMS dataset and choose Analyze/Complex samples/Prepare for Analysis/Create a plan file.
    [Show full text]
  • Choosing Statistical Software
    Choosing Statistical Software Brent Logan, PhD, Professor Mei-Jie Zhang, PhD, Professor Qun Xiang, MS, Biostatistician Medical College of Wisconsin, Division of Biostatistics Friday, June 15, 2012 12:00-1:00 pm Medical Education Building– M2050 The Medical College of Wisconsin is accredited by the Accreditation Council for Continuing Medical Education to provide continuing medical education for physicians. The Medical College of Wisconsin designates this Live activity for a maximum of 1.0 AMA PRA Category 1 Credit(s) ™. Physicians should claim only the credit commensurate with the extent of their participation in the activity. Hours of Participation for Allied Health Professionals The Medical College of Wisconsin designates this activity for up to 1.0 hours of participation for continuing education for allied health professionals. 2 Financial Disclosure • In accordance with the ACCME® standard for Commercial Support Number 6, all in control of content disclosed any relevant financial relationships. The following in control of content had no relevant financial relationships to disclose. Name: Role in Meeting: Ruta Brazauskas, PhD Planning Committee Haley Montsma, BBA Planning Committee Brent Logan, PhD Speaker Mei-Jie Zhang, PhD Speaker Qun Xiang, MS Speaker 3 Learning Objectives • Familiarize the audience with basic statistical software options • Discuss the pros and cons of different types of software • Demonstrate the different types of software on a simple analysis 4 Evaluation Forms Your opinion matters! Help us plan future meetings,
    [Show full text]
  • Analyzing Complex Sample Survey Data Using SAS®, SPSS®, STATA®, and SUDAAN® Arlene B
    NESUG 18 Posters The Big Four: Analyzing Complex Sample Survey Data Using SAS®, SPSS®, STATA®, and SUDAAN® Arlene B. Siller, CDC/National Center for Health Statistics, Hyattsville, MD Linda Tompkins, CDC/National Center for Health Statistics, Hyattsville, MD Abstract The National Center for Health Statistics conducts surveys with probability-based complex sample designs to produce estimates of person health conditions, vital statistics, medical establishment characteristics, disparities in health status and use of health care by race/ethnicity, socio-economic status, and other population traits in the United States. Researchers usually present descriptive statistics such as means, totals, and their standard errors. However, in order to make statistically valid population inferences from sample data, standard errors must be computed using procedures that take into account the complex nature of the sample design. This paper compares estimates produced with four statistical packages (SAS, SPSS, STATA and SUDAAN) using data from two of the many surveys conducted by the Center. Introduction The complex design of sample surveys dictate that the statistical analysis procedures used to analyze the data have the ability to account for multiple stages of sampling, stratification, and clustering. The National Center for Health Statistics conducts such surveys to produce estimates of health conditions and access to medical care for the United States population. These usually include descriptive statistics such as means, totals, and their standard errors. The accompanying standard errors must be computed using procedures that take into account the complex nature of the sample design. In recent years, there has been an increasing effort by statistical software designers to incorporate such procedures.
    [Show full text]