Abstract
We review some issues related to the implications of different missing data mechanisms on statistical inference for contingency tables and consider simulation studies to compare the results obtained under such models to those where the units with missing data are disregarded. We confirm that although, in general, analyses under the correct missing at random and missing completely at random models are more efficient even for small sample sizes, there are exceptions where they may not improve the results obtained by ignoring the partially classified data. We show that under the missing not at random (MNAR) model, estimates on the boundary of the parameter space as well as lack of identifiability of the parameters of saturated models may be associated with undesirable asymptotic properties of maximum likelihood estimators and likelihood ratio tests; even in standard cases the bias of the estimators may be low only for very large samples. We also show that the probability of a boundary solution obtained under the correct MNAR model may be large even for large samples and that, consequently, we may not always conclude that a MNAR model is misspecified because the estimate is on the boundary of the parameter space.
Similar content being viewed by others
References
Baker, S.G., Laird, N.M.: Regression analysis for categorical variables with outcome subject to nonignorable nonresponse. J. Am. Stat. Assoc. 83, 62–69 (1988) (p. 1232, correction)
Baker, S.G., Rosenberger, W.F., DerSimonian, R.: Closed-form estimates for missing counts in two-way contingency tables. Stat. Med. 11, 643–657 (1992)
Brown, C.H.: Protecting against nonrandomly missing data in longitudinal studies. Biometrics 46, 143–156 (1990)
Chen, T.T., Fienberg, S.E.: Two-dimensional contingency tables with both completely and partially cross-classified data. Biometrics 30, 629–642 (1974)
Clarke, P.S.: On boundary solutions and identifiability in categorical regression with non-ignorable non-response. Biom. J. 44, 701–717 (2002)
Clarke, P.S., Smith, P.W.F.: Interval estimation for log-linear models with one variable subject to non-ignorable non-response. J. R. Stat. Soc. B 66, 357–368 (2004)
Clarke, P.S., Smith, P.W.F.: On maximum likelihood estimation for log-linear models with non-ignorable non-response. Stat. Probab. Lett. 73, 441–448 (2005)
Cook, R.D.: Assessment of local influence. J. R. Stat. Soc. B 48, 133–169 (1986)
Fay, R.E.: Causal models for patterns of nonresponse. J. Am. Stat. Assoc. 81, 354–365 (1986)
Fitzmaurice, G.M., Laird, N.M., Zahner, G.E.P.: Multivariate logistic models for incomplete binary responses. J. Am. Stat. Assoc. 91, 99–108 (1996)
Forster, J.J., Smith, P.W.F.: Model-based inference for categorical survey data subject to non-ignorable non-response (with discussion). J. R. Stat. Soc. B 60, 57–79,89–102 (1998)
Glonek, G.F.V.: On identifiability in models for incomplete binary data. Stat. Probab. Lett. 41, 191–197 (1999)
Gustafson, P.: On model expansion, model contraction, identifiability and prior information: two illustrative scenarios involving mismeasured variables (with discussion). Stat. Sci. 20, 111–140 (2005)
Jansen, I., Hens, N., Molenberghs, G., Aerts, M., Verbeke, G., Kenward, M.G.: The nature of sensitivity in monotone missing not at random models. Comput. Stat. Data Anal. 50, 830–858 (2006)
Little, R.J.A., Rubin, D.B.: Statistical Analysis with Missing Data, 2nd ed. Wiley, New York (2002)
Michiels, B., Molenberghs, G.: Protective estimation of longitudinal categorical data with nonrandom dropout. Commun. Stat. Theory Methods 26, 65–94 (1997)
Molenberghs, G., Kenward, M.G.: Missing data in clinical studies. Wiley, New York (2007)
Molenberghs, G., Goetghebeur, E., Lipsitz, S.R., Kenward, M.G.: Nonrandom missingness in categorical data: strengths and limitations. Am. Stat. 53, 110–118 (1999)
Molenberghs, G., Kenward, M.G., Goetghebeur, E.: Sensitivity analysis for incomplete contingency tables: the Slovenian plebiscite case. Appl. Stat. 50, 15–29 (2001)
Molenberghs, G., Beunckens, C., Sotto, C., Kenward, M.G.: Every missingness not at random model has a missingness at random counterpart with equal fit. J. R. Stat. Soc. B 70, 371–388 (2008)
Murray, G.D., Findlay, J.G.: Correcting for the bias caused by drop-outs in hypertension trials. Stat. Med. 7, 941–946 (1988)
Park, T.: An approach to categorical data with nonignorable nonresponse. Biometrics 54, 1579–1590 (1998)
Park, T., Brown, M.B.: Models for categorical data with nonignorable nonresponse. J. Am. Stat. Assoc. 89, 44–52 (1994)
Paulino, C.D.: Analysis of incomplete categorical data: a survey of the conditional maximum likelihood and weighted least squares approaches. Braz. J. Probab. Stat. 5, 1–42 (1991)
Paulino, C.D., Pereira, C.A.B.: Bayesian methods for categorical data under informative general censoring. Biometrika 82, 439–446 (1995)
Rotnitzky, A., Cox, D.R., Bottai, M., Robins, J.M.: Likelihood-based inference with singular information matrix. Bernoulli 6, 243–284 (2000)
Rubin, D.B.: Characterizing the estimation of parameters in incomplete-data problems. J. Am. Stat. Assoc. 69, 467–474 (1974)
Rubin, D.B.: Inference and missing data. Biometrika 63, 581–592 (1976)
Rubin, D.B.: Multiple Imputation for Nonresponse in Surveys. Wiley, New York (1987)
Schafer, J.L.: Analysis of Incomplete Multivariate Data. Chapman & Hall/CRC, Boca Raton (1997)
Smith, P.W.F., Skinner, C.J., Clarke, P.S.: Allowing for non-ignorable non-response in the analysis of voting intention data. Appl. Stat. 48, 563–577 (1999)
Soares, P., Paulino, C.D.: Incomplete categorical data analysis: a Bayesian perspective. J. Stat. Comput. Simul. 69, 157–170 (2001)
Soares, P., Paulino, C.D.: Log-linear models for coarse categorical data. In: Gomes, M.I., Pestana, D., Silva, P. (eds.) Proc. 56th Session of the Internat. Statist. Inst., Invited Paper Meeting #15 (Bayesian Theory and Practice), LVI Bulletin of the Internat. Statist. Inst., Lisbon (2007)
Vansteelandt, S., Goetghebeur, E., Kenward, M.G., Molenberghs, G.: Ignorance and uncertainty regions as inferential tools in a sensitivity analysis. Stat. Sin. 16, 953–979 (2006)
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Poleto, F.Z., Singer, J.M. & Paulino, C.D. Missing data mechanisms and their implications on the analysis of categorical data. Stat Comput 21, 31–43 (2011). https://doi.org/10.1007/s11222-009-9143-x
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11222-009-9143-x