Canada Spends Logo

Explorez notre analyse du budget 2026

Grant: Discovery Grants Program - Individual

Université de Montréal

$19,000.00

Inference in the presence of outliers and missing data

EF 2012-13

Résumé
Item nonresponse occurs in most, if not all surveys. Nonresponse is an important problem in surveys because response rates have been decreasing over time. In most cases, item nonresponse is treated by using some form of imputation. Another difficulty that occurs frequently in surveys (especially in business surveys) is the presence of outliers. Nonresponse generally leads to biased estimators if the respondents and nonrespondents have different characteristics with respect to the variables measured by the survey, whereas outliers potentially lead to very unstable estimators. It becomes necessary to develop techniques that reduce the nonresponse bias in the context of nonresponse and the variance in the context of outliers. Among the possible applications, the developed techniques play a central role in evaluating existing theories and they allow the survey statisticians to provide better estimators. This project will study several aspects of inference in the presence of imputed data and outliers. First, we study the use of the conditional bias of a unit as a systematic way to derive robust estimators in the presence of outliers. Also, the properties of calibrated robust imputation methods will be established. We will propose a new class of imputation methods, called balanced random imputation methods which will lead to efficient estimators because the imputation variance will be reduced, while the distribution of the variables being imputed will be preserved. Other themes of the proposal are the inference in multi-stage sampling and the estimation of relationships in the presence of imputed data.
Programme
Discovery Grants Program - Individual
Emplacement
Quebec, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data

Tel que publié

Application id
327048-2009
Name nom
Haziza, David
Department département
Mathématiques et de statistique
Organization id
63
Institution établissement
Université de Montréal
Province en
Québec
Province fr
Québec
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2012
Competition year année de concours
2009
Award amount
19000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
Discovery grants programs
Group fr
Programmes de subventions à la découverte
Committee code
14
Committee name en
Statistical Sciences
Committee name fr
Statistique
Area of application code
1205
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Mathematical sciences
Area of application fr
Sciences mathématiques
Research subject code
3003
Research subject group en
Statistics and probability
Research subject group fr
Statistique et probabilité
Research subject en
Survey methodology
Research subject fr
Techniques de sondage
Partie
4
Nb partie
5
Application title
Inference in the presence of outliers and missing data
Keyword
NSERC
Application summary
Item nonresponse occurs in most, if not all surveys. Nonresponse is an important problem in surveys because response rates have been decreasing over time. In most cases, item nonresponse is treated by using some form of imputation. Another difficulty that occurs frequently in surveys (especially in business surveys) is the presence of outliers. Nonresponse generally leads to biased estimators if the respondents and nonrespondents have different characteristics with respect to the variables measured by the survey, whereas outliers potentially lead to very unstable estimators. It becomes necessary to develop techniques that reduce the nonresponse bias in the context of nonresponse and the variance in the context of outliers. Among the possible applications, the developed techniques play a central role in evaluating existing theories and they allow the survey statisticians to provide better estimators. This project will study several aspects of inference in the presence of imputed data and outliers. First, we study the use of the conditional bias of a unit as a systematic way to derive robust estimators in the presence of outliers. Also, the properties of calibrated robust imputation methods will be established. We will propose a new class of imputation methods, called balanced random imputation methods which will lead to efficient estimators because the imputation variance will be reduced, while the distribution of the variables being imputed will be preserved. Other themes of the proposal are the inference in multi-stage sampling and the estimation of relationships in the presence of imputed data.