Ignorability conditions for frequentist non parametric analysis of conditional distributions with incomplete data

Shaun Bender, Daniel F. Heitjan

Research output: Contribution to journalArticle

Abstract

Rubin (1976) derived general conditions under which inferences that ignore missing data are valid. These conditions are sufficient but not generally necessary, and therefore may be relaxed in some special cases. We consider here the case of frequentist estimation of a conditional cdf subject to missing outcomes. We partition a set of data into outcome, conditioning, and latent variables, all of which potentially affect the probability of a missing response. We describe sufficient conditions under which a complete-case estimate of the conditional cdf of the outcome given the conditioning variable is unbiased. We use simulations on a renal transplant data set (Dienemann et al.) to illustrate the implications of these results.

Original languageEnglish (US)
Pages (from-to)5252-5264
Number of pages13
JournalCommunications in Statistics - Theory and Methods
Volume46
Issue number11
DOIs
StatePublished - Jun 3 2017

Fingerprint

Ignorability
Parametric Analysis
Incomplete Data
Conditional Distribution
Conditioning
Latent Variables
Missing Data
Partition
Valid
Sufficient
Necessary
Sufficient Conditions
Estimate
Simulation

Keywords

  • Conditional distributions
  • Frequentist analysis
  • Ignorability
  • Incomplete Data
  • Missing Data

ASJC Scopus subject areas

  • Statistics and Probability

Cite this

@article{dfd5128a68cb4a389a0ef2fddeb71899,
title = "Ignorability conditions for frequentist non parametric analysis of conditional distributions with incomplete data",
abstract = "Rubin (1976) derived general conditions under which inferences that ignore missing data are valid. These conditions are sufficient but not generally necessary, and therefore may be relaxed in some special cases. We consider here the case of frequentist estimation of a conditional cdf subject to missing outcomes. We partition a set of data into outcome, conditioning, and latent variables, all of which potentially affect the probability of a missing response. We describe sufficient conditions under which a complete-case estimate of the conditional cdf of the outcome given the conditioning variable is unbiased. We use simulations on a renal transplant data set (Dienemann et al.) to illustrate the implications of these results.",
keywords = "Conditional distributions, Frequentist analysis, Ignorability, Incomplete Data, Missing Data",
author = "Shaun Bender and Heitjan, {Daniel F.}",
year = "2017",
month = "6",
day = "3",
doi = "10.1080/03610926.2015.1099673",
language = "English (US)",
volume = "46",
pages = "5252--5264",
journal = "Communications in Statistics - Theory and Methods",
issn = "0361-0926",
publisher = "Taylor and Francis Ltd.",
number = "11",

}

TY - JOUR

T1 - Ignorability conditions for frequentist non parametric analysis of conditional distributions with incomplete data

AU - Bender, Shaun

AU - Heitjan, Daniel F.

PY - 2017/6/3

Y1 - 2017/6/3

N2 - Rubin (1976) derived general conditions under which inferences that ignore missing data are valid. These conditions are sufficient but not generally necessary, and therefore may be relaxed in some special cases. We consider here the case of frequentist estimation of a conditional cdf subject to missing outcomes. We partition a set of data into outcome, conditioning, and latent variables, all of which potentially affect the probability of a missing response. We describe sufficient conditions under which a complete-case estimate of the conditional cdf of the outcome given the conditioning variable is unbiased. We use simulations on a renal transplant data set (Dienemann et al.) to illustrate the implications of these results.

AB - Rubin (1976) derived general conditions under which inferences that ignore missing data are valid. These conditions are sufficient but not generally necessary, and therefore may be relaxed in some special cases. We consider here the case of frequentist estimation of a conditional cdf subject to missing outcomes. We partition a set of data into outcome, conditioning, and latent variables, all of which potentially affect the probability of a missing response. We describe sufficient conditions under which a complete-case estimate of the conditional cdf of the outcome given the conditioning variable is unbiased. We use simulations on a renal transplant data set (Dienemann et al.) to illustrate the implications of these results.

KW - Conditional distributions

KW - Frequentist analysis

KW - Ignorability

KW - Incomplete Data

KW - Missing Data

UR - http://www.scopus.com/inward/record.url?scp=85013056998&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85013056998&partnerID=8YFLogxK

U2 - 10.1080/03610926.2015.1099673

DO - 10.1080/03610926.2015.1099673

M3 - Article

AN - SCOPUS:85013056998

VL - 46

SP - 5252

EP - 5264

JO - Communications in Statistics - Theory and Methods

JF - Communications in Statistics - Theory and Methods

SN - 0361-0926

IS - 11

ER -