, Volume 20, Issue 3, pp 589–606 | Cite as

Sensitivity analysis for incomplete continuous data

  • Frederico Z. PoletoEmail author
  • Geert Molenberghs
  • Carlos Daniel Paulino
  • Julio M. Singer
Original Paper


Models for missing data are necessarily based on untestable assumptions whose effect on the conclusions are usually assessed via sensitivity analysis. To avoid the usual normality assumption and/or hard-to-interpret sensitivity parameters proposed by many authors for such purposes, we consider a simple approach for estimating means, standard deviations and correlations. We do not make distributional assumptions and adopt a pattern-mixture model parameterization which has easily interpreted sensitivity parameters. We use the so-called estimated ignorance and uncertainty intervals to summarize the results and illustrate the proposal with a practical example. We present results for both the univariate and the multivariate cases.


Identifiability Ignorance interval Missing data Pattern-mixture model Uncertainty interval 

Mathematics Subject Classification (2000)

62F10 62F03 


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. Allison PD (2001) Missing data. Sage, Thousand Oaks Google Scholar
  2. Daniels MJ, Hogan JW (2000) Reparameterizing the pattern mixture model for sensitivity analyses under informative dropout. Biometrics 56:1241–1248 MathSciNetzbMATHCrossRefGoogle Scholar
  3. Daniels MJ, Hogan JW (2007) Missing data in longitudinal studies: strategies for Bayesian modeling and sensitivity analysis. Chapman & Hall, London Google Scholar
  4. Efron B, Gong G (1983) A leisure look at the bootstrap, the jackknife and cross-validation. Am Stat 37:36–48 MathSciNetCrossRefGoogle Scholar
  5. Fitzmaurice G, Davidian M, Verbeke G, Molenberghs G (2008) Longitudinal data analysis. Chapman & Hall, Boca Raton Google Scholar
  6. Glynn RJ, Laird NM, Rubin DB (1986) Selection modeling versus mixture modeling with nonignorable nonresponse (with discussion). In: Wainer H (ed) Drawing inferences from self-selected samples. Erlbaum, Mahwah, pp 115–151 CrossRefGoogle Scholar
  7. Grab EL, Savage IR (1954) Tables of the expected value of 1/X for positive Bernoulli and Poisson variables. J Am Stat Assoc 49:169–177 MathSciNetzbMATHCrossRefGoogle Scholar
  8. Hyndman RJ, Fan Y (1996) Sample quantiles in statistical packages. Am Stat 50:361–365 CrossRefGoogle Scholar
  9. Kenward MG, Goetghebeur E, Molenberghs G (2001) Sensitivity analysis for incomplete categorical data. Stat Model 1:31–48 zbMATHCrossRefGoogle Scholar
  10. Little RJA (1994) A class of pattern-mixture models for normal incomplete data. Biometrika 81:471–483 MathSciNetzbMATHCrossRefGoogle Scholar
  11. Little RJA, Rubin DB (2002) Statistical analysis with missing data, 2nd edn. Wiley, New York zbMATHGoogle Scholar
  12. Little RJA, Wang YX (1996) Pattern-mixture models for multivariate incomplete data with covariates. Biometrics 52:98–111 zbMATHCrossRefGoogle Scholar
  13. Mendenhall W, Lehman JEH (1960) An approximation to the negative moments of the positive binomial useful in life testing. Technometrics 2:227–242 MathSciNetzbMATHCrossRefGoogle Scholar
  14. Molenberghs G, Kenward MG (2007) Missing data in clinical studies. Wiley, New York CrossRefGoogle Scholar
  15. Molenberghs G, Kenward MG, Goetghebeur E (2001) Sensitivity analysis for incomplete contingency tables: the Slovenian plebiscite case. Appl Stat 50:15–29 zbMATHCrossRefGoogle Scholar
  16. Poleto FZ, Paulino CD, Molenberghs G, Singer JM (2010) Inferential implications of over-parameterization: a case study in incomplete categorical data. Tech rep, RT-MAE-2010-04, Instituto de Matemática e Estatística, Universidade de São Paulo Google Scholar
  17. Rotnitzky A, Robins JM, Scharfstein DO (1998) Semiparametric regression for repeated outcomes with nonignorable nonresponse. J Amer Stat Assoc 93:1321–1339 MathSciNetzbMATHCrossRefGoogle Scholar
  18. Rotnitzky A, Scharfstein D, Su TL, Robins JM (2001) Methods for conducting sensitivity analysis of trials with potentially nonignorable competing causes of censoring. Biometrics 57:103–113 MathSciNetzbMATHCrossRefGoogle Scholar
  19. Rubin DB (1976) Inference and missing data. Biometrika 63:581–592 MathSciNetzbMATHCrossRefGoogle Scholar
  20. Rubin DB (1977) Formalizing subjective notions about the effect of nonrespondents in sample surveys. J Am Stat Assoc 72:538–543 zbMATHCrossRefGoogle Scholar
  21. Scharfstein DO, Rotnitzky A, Robins JM (1999) Adjusting for nonignorable drop-out using semiparametric nonresponse models (with discussion). J Am Stat Assoc 94:1096–1146 MathSciNetzbMATHCrossRefGoogle Scholar
  22. Scharfstein DO, Daniels MJ, Robins JM (2003) Incorporating prior beliefs about selection bias into the analysis of randomized trials with missing outcomes. Biostatistics 4:495–512 zbMATHCrossRefGoogle Scholar
  23. Sen PK, Singer JM, Pedroso de Lima AC (2009) From finite sample to asymptotic methods in statistics. Cambridge University Press, Cambridge Google Scholar
  24. Silverman BW (1986) Density estimation, 2nd edn. Chapman & Hall, London zbMATHGoogle Scholar
  25. Stephan FF (1945) The expected value and variance of the reciprocal and other negative powers of a positive Bernoullian variate. Ann Math Stat 16:50–61 MathSciNetzbMATHCrossRefGoogle Scholar
  26. Vansteelandt S, Goetghebeur E, Kenward MG, Molenberghs G (2006) Ignorance and uncertainty regions as inferential tools in a sensitivity analysis. Stat Sin 16:953–979 MathSciNetzbMATHGoogle Scholar

Copyright information

© Sociedad de Estadística e Investigación Operativa 2010

Authors and Affiliations

  • Frederico Z. Poleto
    • 1
    Email author
  • Geert Molenberghs
    • 2
    • 3
  • Carlos Daniel Paulino
    • 4
  • Julio M. Singer
    • 1
  1. 1.Instituto de Matemática e EstatísticaUniversidade de São PauloSão Paulo, SPBrazil
  2. 2.I-BioStat, Universiteit HasseltDiepenbeekBelgium
  3. 3.Katholieke Universiteit LeuvenLeuvenBelgium
  4. 4.Instituto Superior TécnicoUniversidade Técnica de Lisboa (and CEAUL-FCUL)LisboaPortugal

Personalised recommendations