Advertisement

The use of weights to account for non-response and drop-out

  • Michael HöflerEmail author
  • Hildegard Pfister
  • Roselind Lieb
  • Hans-Ulrich Wittchen
ORIGINAL PAPER

Abstract

Background

Empirical studies in psychiatric research and other fields often show substantially high refusal and drop-out rates. Non-participation and drop-out may introduce a bias whose magnitude depends on how strongly its determinants are related to the respective parameter of interest.

Methods

When most information is missing, the standard approach is to estimate each respondent’s probability of participating and assign each respondent a weight that is inversely proportional to this probability. This paper contains a review of the major ideas and principles regarding the computation of statistical weights and the analysis of weighted data.

Results

A short software review for weighted data is provided and the use of statistical weights is illustrated through data from the EDSP (Early Developmental Stages of Psychopathology) Study. The results show that disregarding different sampling and response probabilities can have a major impact on estimated odds ratios.

Conclusions

The benefit of using statistical weights in reducing sampling bias should be balanced against increased variances in the weighted parameter estimates.

Key words

selection bias non-response drop-out missing values weighting survey 

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    Allehoff WH, Esser G, Schmidt MH, Hennicke K (1983) Die Bedeutung der Informations- und Kooperationsverweigerung für die Interpretationsreichweite einer mehrstufigen kinderpsychiatrisch-epidemiologischen Untersuchung. Soc Psychiatry 18:29–36Google Scholar
  2. 2.
    Binder DA (1983) On the variances of asymptotically normal estimators from complex surveys. Int Stat Rev 51:279–292Google Scholar
  3. 3.
    Brogan DJ (1998) Pitfalls of using standard statistical software packages for sample survey data. In: Armitage P, Colton T (eds) Encyclopedia of Biostatistics, 4167–4174. New York: John Wiley and SonsGoogle Scholar
  4. 4.
    Carpenter J, Bithell J (2000) Bootstrap confidence intervals: when, which, what? A practical guide for statisticians. Stat Med 19:1141–1164Google Scholar
  5. 5.
    Cox DR, Wermuth N (1996) Multivariate dependencies, Chapman und Hall, LondonGoogle Scholar
  6. 6.
    Diggle PJ, Liang K-Y, Zeger SL (1994) Analysis of longitudinal data, Oxford University Press, OxfordGoogle Scholar
  7. 7.
    Efron B, Tibshirani RJ (1993) An introduction to the bootstrap, Chapman and Hall, LondonGoogle Scholar
  8. 8.
    Greenland S (1977) Response and follow-up bias in cohort studies. Am J Epidemiol 106:184–187Google Scholar
  9. 9.
    Greenland S (in press) Multiple-bias modeling for analysis of observational data. J Roy Stat SocGoogle Scholar
  10. 10.
    Heeringa SG,Liu J (1998) Complex sample design effects and inference for mental health survey data. Int J Methods Psychiatr Res 7:56–65Google Scholar
  11. 11.
    Heyting A, Tolboom JTBM, Essers JGA (1992) Statistical handling of dropouts in longitudinal clinical trials. Stat Med 11:2043–2061Google Scholar
  12. 12.
    Insightful Corp. (2003) Documentation for S-PLUS 6.2, Seattle, WA: Insightful CorpGoogle Scholar
  13. 13.
    Jacobi F, Wittchen HU, Holting C, Sommer S, Lieb R, Höfler M, Pfister H (2002) Estimating the prevalence of mental and somatic disorders in the community: aims and methods of the German National Health—Interview and Examination Survey. Int J Methods Psychiatr Res 11:1–18Google Scholar
  14. 14.
    Kauermann G, Carroll RJ (2001) A note on the efficiency of sandwich covariance matrix estimation. J Am Stat Ass 96:1387–1396Google Scholar
  15. 15.
    Kessler RC, McGonagle KA, Zhao S, Nelson CB, Hughes M, Eshleman S, Wittchen HU, Kendler KS (1994) Lifetime and 12-month prevalence of DSM-III-R psychiatric disorders in the United States: Results from the National Comorbidity Survey. Arch Gen Psychiatr 51:8–19Google Scholar
  16. 16.
    Kessler RC, Little RJA, Groves RM (1995) Advances in strategies for minimizing and adjusting for survey nonresponse. Epidem Rev 17:192–204Google Scholar
  17. 17.
    Kish L, Frankel MR (1970) Balanced repeated replications for standard errors. J Amer Stat Ass 65:1071–1095Google Scholar
  18. 18.
    Levy PS, Lemeshow S (1999) Sampling of populations—methods and application, John Wiley and Sons, New YorkGoogle Scholar
  19. 19.
    Lieb R, Isensee B, Von Sydow K, Wittchen HU (2000) The Early Developmental Stages of the Psychopathology Study (EDSP): A methodological update. Eur Add Res 6:170–182Google Scholar
  20. 20.
    Light RJ, Singer JD, Willett JB (1990) By design—planning research on higher education. Harvard University Press, Cambridge MAGoogle Scholar
  21. 21.
    Little RJA (1986) Survey nonresponse adjustments for estimates of means. Int Stat Rev 54:139–157Google Scholar
  22. 22.
    Little RJA, Lewitzky S, Heeringa S, Lepkowski J, Kessler RC (1997) Assessment of weighted methodology for the national comorbidity survey. Am J Epidem 146:439–449Google Scholar
  23. 23.
    McCullagh P, Nelder JA (1989) Generalized Linear Models, 2nd edition. Chapman and Hall, LondonGoogle Scholar
  24. 24.
    Miller ME, Ten Have TR, Reboussin BA, Lohmann KK, Rejeski WJ (2001) A marginal model for analysing discrete outcomes from longitudinal surveys with outcomes subject to multiple-cause non-response. J Am Stat Ass 95:844–857Google Scholar
  25. 25.
    Pigeot I (2001) The jackknife and bootstrap in biomedical research—Common principles and possible pitfalls. Drug Information Journal 35:1431–1443Google Scholar
  26. 26.
    Preisser JS, Galecki AT, Lohmann KK, Wagenknecht LE (2000) Analysis of smoking trends with incomplete longitudinal binary responses. J Am Stat Ass 95:1021–1031Google Scholar
  27. 27.
    Rao JNK, Shao J (1999) Modified balanced repeated replication for complex survey data. Biometrika 86:403–415Google Scholar
  28. 28.
    Rotnitzky A, Robins J (1997) Analysis of semi-parametric regression models with non-ignorable non-response. Stat in Med 16:81–102Google Scholar
  29. 29.
    Rosenbaum PR (2002) Observational Studies, 2nd edition Springer, New YorkGoogle Scholar
  30. 30.
    Rosenbaum PR,Rubin DB (1983) The central role of the propensity score in observational studies for causal effects. Biometrika 70:41–55Google Scholar
  31. 31.
    Rubin DB (1996) Multiple imputation after 18+ years. J Amer Stat Ass 91:473–489Google Scholar
  32. 32.
    Rubin DB (2003) Discussion on multiple imputation. Int Stat Rev 71:619–625Google Scholar
  33. 33.
    Rubin DB, Schenker N (1991) Multiple imputation in health care databases: an overview and some applications. Stat Med 10:585–598Google Scholar
  34. 34.
    Royall RM (1986) Model robust confidence intervals using maximum likelihood estimators. Int Stat Rev 54:221–226Google Scholar
  35. 35.
    Schafer JL (1997) Analysis of incomplete multivariate data, Chapman and Hall, LondonGoogle Scholar
  36. 36.
    Scharfstein DO, Rotnitzky A, Robins JM (1999) Adjusting for nonignorable dropout using semiparametric nonresponse models. J Am Stat Ass 94:1096–1120Google Scholar
  37. 37.
    Smith TMF (2001) Biometrika Centenary: sample surveys. Biometrika 88:167–194Google Scholar
  38. 38.
    SAS Institute Inc. (2003) SAS OnlineDoc® 9.1. Cary, NC: SAS Institute IncGoogle Scholar
  39. 39.
    Shah BV, Barnwell BG, Bieler GS (2004). SUDAAN User’s manual: Release 9.0, NC: Research Triangle Institute, Research Triangle ParcGoogle Scholar
  40. 40.
    SPSS Inc. (2004) SPSS for Windows Version 13. Chicago, IL: SPSS IncGoogle Scholar
  41. 41.
    StataCorp. Stata Statistical Software: Release 8.0 (2003) College Station, TX: Stata CorporationGoogle Scholar
  42. 42.
    Touloumi G, Pocock SJ, Babiker AG, Darbyshire JH (2002) Impact of missing data due to selective dropouts in cohort studies and clinical trials. Epidem 13:347–355Google Scholar
  43. 43.
    Troxel AB, Lipsitz SR, Brennan TA (1997) Weighted estimation equations with nonignorably missing response data. Biometrics 53:857–869Google Scholar
  44. 44.
    White H (1982) Maximum likelihood estimation of misspecified models. Econometrica 50:1–25Google Scholar
  45. 45.
    Wittchen H, Höfler M, Gander F, Pfister H, Storz S, Üstün TB, Miller N, Kessler RC (1999) Screening for mental disorders: performance of the Composite International Diagnostic-Screener (CID-S). Int J Meth Psychiatr Res 8:59–70Google Scholar
  46. 46.
    Wittchen HU, Perkonigg A, Lachner G, Nelson CB (1998a) Early Developmental Stages of Psychopathology Study (EDSP): Objectives and design. Eur Add Res 4:18–27Google Scholar
  47. 47.
    Wittchen HU, Nelson CB, Ladner G (1988b) Prevalence of mental disorders and psychosocial impairments in adolescents and young adults. Psychol Med 28:109–126Google Scholar
  48. 48.
    Yung W, Rao JNK (2000) Jackknife variance estimation under imputation for estimators using poststratification information. J Am Stat Ass 903–915Google Scholar

Copyright information

© Steinkopff Verlag 2005

Authors and Affiliations

  • Michael Höfler
    • 1
    Email author
  • Hildegard Pfister
    • 1
  • Roselind Lieb
    • 1
  • Hans-Ulrich Wittchen
    • 1
    • 2
  1. 1.Max-Planck-Institut of Psychiatry, Clinical Psychology and EpidemiologyMünchenGermany
  2. 2.Technical University Dresden, Institute of Clinical Psychology & PsychotherapyDresdenGermany

Personalised recommendations