DOI QR코드

DOI QR Code

Composite estimation type weighting adjustment for bias reduction of non-continuous response group in panel survey

패널조사에서 비연속 응답 그룹 편향 보정을 위한 복합가중값

  • Choi, Hyunga (Employment Statistics Survey Team, Korea Employment Information Service) ;
  • Kim, Youngwon (Department of Statistics, Sookmyung Women's University)
  • 최형아 (한국고용정보원 고용통계조사팀) ;
  • 김영원 (숙명여자대학교 통계학과)
  • Received : 2019.04.09
  • Accepted : 2019.04.23
  • Published : 2019.06.30

Abstract

Sample attrition according to a long-term tracking reduces the representativeness of the sample data in a panel study. Most panel surveys in South Korea and other countries have prepared response adjustment weights in order to solve problems regarding representativeness due to sample attrition. In this paper, we divided the panel data into continuous response group and non-continuous response group according to response patterns and considered a weighting adjustment method to reduce the bias of the non-continuous response group. A simulation indicated that the proposed composite estimation type weighting method, which reflected the characteristics of non-continuous response groups, could be more efficient than other weighting methods in terms of reducing non-response bias. As a case study, the proposed methods are applied to the Korean Longitudinal Study of Ageing (KLoSA) data of the Korea Employment Information Service.

패널 자료는 자료가 축적되는 만큼 그 가치가 증대된다. 이와 동시에 장기추적에 따른 표본이탈은 자료의 신뢰성을 떨어뜨린다. 국내 외 대부분의 패널조사에서 가중값 보정을 통해 표본 이탈 문제를 해결하고 있다. 본 논문에서는 패널자료에서 차수별 응답여부에 따라 연속 응답 그룹과 비연속 응답 그룹으로 나누고, 비연속 응답 그룹에 대한 적정 가중값 산출방법을 검토하였다. 연속/비연속 응답그룹을 구분하여 비연속 응답 그룹의 응답자 특성을 반영한 복합추정 방식의 가중값 작성방법을 제안하고, 그룹의 구분 없이 작성하였던 기존의 가중값 작성방법과 새로 제안한 복합추정 방식의 가중값 산출방법의 효율성을 모의실험과 실증분석을 통해 살펴보았다. 결과적으로 새로 제안한 복합추정 방식의 가중값 산출방법은 기존 방법 보다 편향을 대폭 감소시킴을 모의실험을 통해 볼 수 있었다. 한편, 제시한 가중값 작성방법을 한국고용정보원 고령화연구패널에 적용한 결과도 제시하였다.

Keywords

GCGHDE_2019_v32n3_375_f0001.png 이미지

Figure 2.1. Continuous and non-continuous response groups at wave t.

GCGHDE_2019_v32n3_375_f0002.png 이미지

Figure 3.1. Wave t response rate versus wave 2 response rate by group.

GCGHDE_2019_v32n3_375_f0003.png 이미지

Figure 3.2. Difference of mean by continuous and non-continuous response group.

GCGHDE_2019_v32n3_375_f0004.png 이미지

Figure 3.3. Scatter plot of direct estimate (p_b) and composite estimate (p_b) for response rate of non-continuous response group (Data: KLoSA, KEIS).

GCGHDE_2019_v32n3_375_f0005.png 이미지

Figure 5.1. Comparison of bias with extreme weight (above) and bias after treatment of extreme weights (below).

GCGHDE_2019_v32n3_375_f0006.png 이미지

Figure 5.2. Bias of method 5 and method 6 with response propensity variable zi.

Table 4.1. Scenario of three-year panel data for simulation

GCGHDE_2019_v32n3_375_t0001.png 이미지

Table 4.2. Generating variable of interest (yi) for simulation

GCGHDE_2019_v32n3_375_t0002.png 이미지

Table 4.3. Simulation result

GCGHDE_2019_v32n3_375_t0003.png 이미지

Table 5.1. Weight statistic by method (bias of annual personal gross income)

GCGHDE_2019_v32n3_375_t0004.png 이미지

References

  1. Baek, J. S. and Shim, K. S. (2012). How to create cross weights in household panel survey, 2012 Second-Half Research Report 3, National Bureau of Korea.
  2. Brady, T. W. (2013). The effects of errors in paradata on weighting class adjustments: a simulation study, Improving Survey with Paradata: Analytic Uses of Process Information, John Wiley & Sons Inc.
  3. Josiane, B., David, H., Anni, O., and Melanie, N. (2018). The Dynamics of Ageing : Evidence from the English Longitudinal Study of Ageing 2002-16(Wave 8) Ch5. Methodology, National Center for Social Research.
  4. Kim, Y. W., Lee, K. J., and Park, I. H. (2015). Report of Recreating Weights for KLoSA Wave 1-4 and Weights for Wave 5, Korea Employment Information Service.
  5. Korea Institute of Public Finance (2012). Weights for National Survey of Tax and Benefit (Technical Report), Korea.
  6. Lee, J. R., Choi, E. Y., Do, N. H., Song, S. Y., Wang, Y. H., and Jung, Y. H. (2011). Panel Study on Korean Children 2011 Report, Korea Institute of Child Care and Education.
  7. Park, M. G. and Kim, Y. W. (2013). Research Report of YP Wave 6 and New Response Model Development, Korea Employment Information Service.
  8. Park, M. G., Kim, Y. W., and Byun, J. S. (2013). Weights Research of Korean Labor & Income Panel Survey, Korea Labor Institute.
  9. Park, M. G., Lee, K. S., Park, H. S., and Kang, H. C. (2011). A Study on the Construction of Weights for KYPS. Survey Research 12(3) 173-186. The Korean Association for Survey Research.
  10. Qixuan, C., Andrew, G., Melissa, T., Fran, H. N. and Sandro, G. (2012). Weighting Adjustments for Panel Nonresponse. Unpublished manuscript. Columbia University.
  11. Sally, B., David, H., and Margaret, B. (2015). The Dynamics of Aging : The 2012 English Longitudinal Study of Ageing(Wave 6). National Center for Social Research.
  12. Shaun, S., Kate, C., and Carli, L. (2008). Living in the 21st century : older people in England the 2006 English Longitudinal Study of Ageing (Wave 3) Ch9. Methodology. National Center for Social Research.
  13. Shin, J. G., Hwang, K. H., and Cho, M. S. (2017). YP2007 Wave10 Analysis Report, Korea Employment Information Service.
  14. Takis, M. (2010). Combining information from multiple surveys by using regression for efficient small domain estimation, Journal of the Royal Statistical Society. Series B (Statistical Methodology), 72, 27-48. https://doi.org/10.1111/j.1467-9868.2009.00724.x