Contents

---

Introduction

Data Collection and Processing Procedures

Coverage Issues

Statistical Issues

Summary and Conclusions

Appendix A. SESTAT and CPS Procedures for Collecting/Processing Data

Appendix B. SESTAT and CPS Variables Used

Appendix C. Standard Errors of SESTAT and CPS Estimates

Appendix D. CPS Individuals in S&E Occupations, by Highest Degree Attained

Appendix E. CPS Individuals in S&E Occupations With Degree Below Bachelor's

References

Suggested Citation, Acknowledgments

Appendix C. Standard Errors of SESTAT and CPS Estimates

Computational Procedures

The procedures used to compute the approximate variance of the SESTAT and CPS estimates presented in this report are based on procedures described in two technical reports: Cox et al. (2000) and U.S. Census Bureau (2000). In both reports, similar generalized variance function (GVF) models are used to approximate standard errors for estimates of characteristics of the target population. Using the GVF, the standard error (SE) of an estimated total can be obtained by inserting the value of the estimated total, , into the following formula:

    (C.1)

where and are appropriate constants.

The coefficient of variation (or relative standard error) of an estimated total is defined as the standard error of the estimate divided by the estimate, that is,

  .    (C.2)

Functions of the form (C.1) and (C.2) are referred to as GVFs and are derived empirically using standard errors computed for a large number of survey estimates. For the SESTAT data, GVFs have been developed for the overall population and for subgroups defined by sex, race/ethnicity, and occupation. To obtain approximate standard errors of estimates presented in this report, see Cox et al. (2000), appendix B, tables B-1 to B-4. To obtain the standard errors of the corresponding CPS estimates, the GVF provided in U.S. Department of Labor, Bureau of Labor Statistics (1997) is used.

Because the two survey systems are independent, the standard error of a difference between CPS and SESTAT estimates, , can be computed as

   .    (C.3)

 Back to Top

Standard Errors of Estimated Totals and Differences

The following tables summarize the estimates, standard errors, CVs, and corresponding sample sizes of the estimates presented in the section "Coverage Issues" of this report. Tables C-1 to C-8 correspond to SESTAT estimates, tables C-9 to C-28 correspond to CPS estimates, and tables C-29 to C-32 and figure C-1 correspond to differences between the SESTAT and CPS estimates.