Abstract
We propose a new method to make inference about an eye parameter based on data that contains measurements on both eyes for some subjects, and measurements on only one eye on the others. Subject effects are modeled as additive and random, and correlation between observations on the same subject are taken into account. We derive confidence intervals for the parameter of interest, using unbiased estimators of mean and variance, and yielding explicit formulas. The results are compared to approaches commonly used in the literature, using theoretical considerations and a simulation study. The method works well even for small to moderate sample sizes, for continuous and for discrete data, and it is applicable generally for the situation where data is collected partially in pairs and partially in singles, and inference is to be made about a common location parameter. The conclusions as to how one should best average correlated data may be surprising, and somewhat revising conventional wisdom.
Similar content being viewed by others
References
Agresti A, Caffo B (2000) Simple and effective confidence intervals for proportions and differences of proportions result from adding two successes and two failures. Am Stat 54(4):280–288
Altman DG, Martin Bland J (1997) Units of analysis. Br Med J 314:1874
Armstrong RA (2013) Statistical guidelines for the analysis of data obtained from one or both eyes. Ophthalmic Physiol Opt 33:7–14
Bates D, Maechler M, Bolker B, Walker S, Christensen RHB, Singmann H, Dai B (2014) R Package lme4, version 1.1–7. http://CRAN.R-project.org/package=lme4
Bland JM, Altman DG (1997) Correlation, regression, and repeated data. Br Med J 314:1874
Bryant D, Havey TC, Roberts R, Guyatt G (2006) How many patients? How many limbs? Analysis of patients or limbs in the orthopaedic literature: a systematic review. J Bone Jt Surg 88:41–45
Fan Q, Teo Y-Y, Saw S-M (2011) Application of advanced statistics in ophthalmology. IOVS 52(9):6059–6065
Gøtzsche PC (1989) Methodology and overt and hidden bias in reports of 196 double-blind trials of nonsteroidal antiinflammatory drugs in rheumatoid arthritis. Control Clinidal Trials 10:31–56
Holopigian K, Bach M (2010) A primer on common statistical errors in clinical ophthalmology. Doc Ophthalmol 121:215–222
ICH E9 Expert Working Group (1998) ICH harmonised tripartite guideline: statistical principles for clinical trials. Stat Med 18(15):1903–1904
Karakosta A, Vassilaki M, Plainis S, Elfadl NH, Tsilimbaris M, Moschandreas J (2012) Choice of analytic approach for eye-specific outcomes: one eye or two? Am J Ophthalmol 153:571–579
Katz J, Zeger S, Liang K-Y (1994) Appropriate statistical methods to account for similarities in binary outcomes between fellow eyes. Investig Ophthalmol Vis Sci 35(5):2461–2465
Konietschke F, Harrar SW, Lange K, Brunner E (2011) Ranking procedures for matched pairs with missing data—asymptotic theory and a small sample approximation. Comput Stat Data Anal 56:1090–1102
Mangaleshkar SR, Rajesh MB, Tandon RK (2001) Surgical correction of severe claw toe deformity: a review of the Stainsby procedure. Foot 11:126–131
Menz HB (2004) Two feet, or one person? Problems associated with statistical analyses of paired data in foot and ankle medicine. Foot 14:2–5
Menz HB (2005) Analysis of paired data in physical therapy research: time to stop double-dipping? J Orthopaedic Sports Phys Ther 35(8):477–478
Murdoch IE, Morris SS, Cousens SN (1998) People and eyes: statistical approaches in ophthalmology. Br J Ophthalmol 82:971–973
Nadarajah S, Kotz S (2008) Exact distribution of the max/min of two gaussian random variables. IEEE Trans Very Large Scale Integr (VLSI) Syst 16(2):210–212
Newcombe RG, Duff GR (1987) Eyes or patients? Traps for the unwary in the statistical analysis of ophthalmological studies. Br J Ophthalmol 71:645–646
R Core Team (2014) R: a language and environment for statistical computing. R Foundation for Statistical Computing, Wien, Austria. http://www.R-project.org
Ricketti JC (2005) Terbinafine/miconazole nitrate 2% tincture compound for the treatment of onychomycosis. Foot 11:21–23
Rosner B (1982) A note on statistical methods adjusting for intraclass correlation. Biometrics 38(1):105–114
Rosner B (1984) Multivariate methods in ophthalmology with application to other paired-data situations. Biometrics 40(4):1025–1035
Rubin DB (1976) Inference and missing data. Biometrika 63(3):581–592
Sutton AJ, Muir KR, Jones AC (1997) Two knees or one person: data analysis strategies for paired joints or organs. Ann Rheum Dis 56:401–402
Velu R, McInerney M (1985) A note on statistical methods adjusting for intraclass correlation. Biometrics 41:533–538
Venables WN, Ripley BD (2002) Modern applied statistics with S, 4th edn. Springer, New York
Xu J, Harrar SW (2012) Accurate mean comparisons for paired samples with missing data: an application to a smoking-cessation trial. Biom J 54(2):281–295
Acknowledgments
We thank the Department of Ophthalmology, Konventhospital Barmherzige Brueder Linz, for providing anonymized eye- and subject-specific data sets. Furthermore, we would like to thank the editorial team and the reviewers, as well as our colleagues Edgar Brunner, Ludwig Hothorn, and Larry Madden, for helpful comments and excellent suggestions.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Fuchs, N., Pölz, W. & Bathke, A.C. Confidence intervals for population means of partially paired observations. Stat Papers 58, 35–51 (2017). https://doi.org/10.1007/s00362-015-0686-y
Received:
Revised:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00362-015-0686-y