Inference and Missing Data.¶
RUBIN, D. B. (1976). Inference and Missing Data. Biometrika, 63(3), 581-592.
Cited by¶
11 citations across 7 artifacts.
Each citation links to the sentence it supports in the citing article.
Primes¶
- Imputation
- The fill is applied with an explicit assumption about the relationship between observed and missing — missing-completely-at-random, missing-at-random, or missing-not-at-random, in the formal idiom.
This sourceIntroduces the MCAR/MAR/MNAR missingness taxonomy and the conditions under which the missingness mechanism can be ignored — the load-bearing assumption framework for imputation, including the MNAR-blindness result.
Supported in partVerified against the publisher's abstract
Rubin's abstract states the MAR / 'observed at random' ignorability conditions, not the MCAR/MAR/MNAR triad the claim names.
“it is appropriate to ignore the process that causes missing data if the missing data are ‘missing at random’ and the observed data are ‘observed at random’”
- The fill is applied with an explicit assumption about the relationship between observed and missing — missing-completely-at-random, missing-at-random, or missing-not-at-random, in the formal idiom.
- Missing Data Mechanisms (MCAR, MAR, MNAR)
- (1) **Missing data mechanisms classify the process by which observations become missing into three increasingly problematic categories: MCAR (missing completely at random)
This sourceRubin foundational taxonomy of missing-completely-at-random (MCAR).
Supported in partVerified against the publisher's abstract
Rubin's abstract gives the conditions (missing at random, observed at random, distinctness) under which the missing-data process may be ignored, which underlies MCAR/MAR, but it names no three-category taxonomy and no ordering of severity.
“it is appropriate to ignore the process that causes missing data if the missing data are ‘missing at random’ and the observed data are ‘observed at random’, but these inferences are generally conditional on the observed pattern of missing data”
- where missingness is independent of all variables observed and unobserved; MAR (missing at random)
This sourceRubin missing-at-random (MAR) category in foundational taxonomy.
Supported in partVerified against the publisher's abstract
Rubin's abstract names 'missing at random' as his own condition for ignoring the missing-data process, backing the MAR half of the claim but not the MCAR independence definition.
“it is appropriate to ignore the process that causes missing data if the missing data are ‘missing at random’ and the observed data are ‘observed at random’”
- where missingness depends only on observed variables; and MNAR (missing not at random)
This sourceRubin missing-not-at-random (MNAR) category in foundational taxonomy.
Supported in partVerified against the publisher's abstract
Rubin's abstract supplies the foundational 'missing at random' frame, but not the clause that missingness depends only on observed variables, nor the MNAR category.
“it is appropriate to ignore the process that causes missing data if the missing data are ‘missing at random’ and the observed data are ‘observed at random’”
- … where missingness is independent of all variables observed and unobserved; MAR (missing at random) where missingness depends only on observed variables; and MNAR (missing not at random) where missingness depends on the unobserved values themselves.** (2) The classification, formalized by Donald Rubin in 1976
Supported in partVerified against the publisher's abstract
The abstract establishes Rubin (1976) as a foundational treatment of missing at random and conditions for ignoring the missing-data process, but it does not state the MCAR/MAR/MNAR classification or its formalization.
“it is appropriate to ignore the process that causes missing data if the missing data are ‘missing at random’ and the observed data are ‘observed at random’”
- Donald Rubin's 1976 paper "Inference and Missing Data" (Biometrika)
This sourceRubin foundational missing-data taxonomy paper.
SupportedVerified against the publisher's abstract
The cited work is Rubin's "Inference and missing data", in Biometrika for 1976, and its summary treats inference where the missing-data process matters, working with data "missing at random", "observed at random", and parameters that are "distinct". Author, year, title and journal are as the sentence gives them. The work's own summary does not restate its title, year or journal; those match through the publication record rather than through its prose.
“Inference and missing data”
- (1) **Missing data mechanisms classify the process by which observations become missing into three increasingly problematic categories: MCAR (missing completely at random)
- Selection Bias
- The 1936 Literary Digest failure (2.4 million respondents predicting the wrong election outcome) is the classic demonstration
This sourceRubin foundational taxonomy of missing-completely-at-random (MCAR).
- The 1936 Literary Digest failure (2.4 million respondents predicting the wrong election outcome) is the classic demonstration
- Selection-Visibility Gate
- Dropout uncorrelated with expression is attrition, not a gate.
This sourceSeparates missingness independent of the underlying value from missingness that depends on it, which is what distinguishes ordinary attrition from a selective gate.
Borderline supportVerified against the publisher's abstract
Rubin's condition states that missingness independent of the underlying value may be ignored, which is the attrition-versus-gate distinction the claim draws.
“it is appropriate to ignore the process that causes missing data if the missing data are ‘missing at random’ and the observed data are ‘observed at random’”
Two runs of the same check disagreed on this citation; the later verdict is shown and the citation is queued for review.
- Dropout uncorrelated with expression is attrition, not a gate.
- Silence as Signal
- The cost-asymmetric mechanism is precisely what makes the pattern substrate-independent, operating in physics through detection thresholds, in biology through sampling effort, in social systems through reporting cost, and in digital systems through logging asymmetries; a practitioner who has corrected for it once carries both the diagnostic and the four-family repair into every record that was expensive to produce and cheap to omit.
This sourceFoundational taxonomy of missing-data mechanisms (MCAR, MAR, MNAR); silence-as-signal is the missing-not-at-random case with a cost-asymmetric mechanism.
Supported in partVerified against the publisher's abstract
Rubin's abstract establishes that missing-data processes can be non-ignorable and must be modelled, but never states a cost-asymmetric mechanism or substrate-independence across four domains.
“These conditions are the weakest general conditions under which ignoring the process that causes missing data always leads to correct inferences.”
- The cost-asymmetric mechanism is precisely what makes the pattern substrate-independent, operating in physics through detection thresholds, in biology through sampling effort, in social systems through reporting cost, and in digital systems through logging asymmetries; a practitioner who has corrected for it once carries both the diagnostic and the four-family repair into every record that was expensive to produce and cheap to omit.
Mechanisms¶
- Data Completeness Check
- Separating declared-missing from silently-missing is the entire point of the exercise, and it draws on the standard vocabulary of missing-data mechanisms.
This sourceIntroduces the standard taxonomy of missing-data mechanisms according to how missingness relates to observed and unobserved values.
- Separating declared-missing from silently-missing is the entire point of the exercise, and it draws on the standard vocabulary of missing-data mechanisms.
- Union Delta Review
Verification¶
Does it exist? Confirmed. This work's DOI resolves to a registered record, which fixes its identity. That is all it fixes.
Does it back the claim? Read against the text for 8 of 11 citations: 1 supported, 6 supported in part, 1 borderline. Each verdict is shown under its citation below, with what in the work backs the sentence.
Support is checked per citation rather than per work — the same source can be cited soundly in one article and wrongly in another. Per-citation recording began recently, so a citation with no recorded check is a gap in the record rather than evidence it went unchecked.
See how references were verified.
Links previously used in the corpus¶
Before the registry existed this work was also linked 1 other way.
Registry ID ref:1d90c3160b4e · see in the full table