Grant: Discovery Grants Program - Individual
McGill University
$28,000.00
Statistical inference with parameter indeterminacies, robustness and causality
EF 2016-17
Résumé
In this program, I intend to pursue research in three wide related areas:(A) statistical inference in the presence of identification problems;(B) robustness to model incompleteness(C) causality analysis in time series(A) In the first area, I wish to pursue my earlier work on the statistical problems associated withobservationally equivalent parameter vectors (identification) which often occur in many statistical models, for example structural models derived from economic theory which suggest that explanatory variables in a regression should be viewed as ""endogenous"". Even though it is commonly accepted that standard inference procedures (estimators, tests, confidence sets) become invalid in such models, I wish to take a contrarian view and explore in a systematic way when and in what ways standard inference procedures not designed for structural models [such as ordinary least squares estimators, maximum likelihood, various testing procedures, etc] remain ""valid"" in such models (if not possibly superior to methods designed to deal with such models) - with or without ""strong identification"" - to what extent forecasts based on such methods can still yield useful results, etc. This research may require redefining somewhat what is meant by a valid inference. Related problems that will be studied include:(1) the relationship with set identification and testability;(2) instrument selection, taking into account the fact that methods specifically designed for inference on structural models (such as IV methods) may be inferior in finite samples to methods that do not take ""endogeneity"" into account; (3) applications to factor analysis; (4) applications to inference on dynamic general equilibrium models.(B) In the second area, I wish to study the statistical theory associated with inference on incompletely specified models: what restrictions are needed to ensure that a statistical procedure remains valid despite the fact that the process governing some variables is not specified (or is incompletely specified)? This issue will be explored by using methods based on the theory of invariant statistical procedures, sufficient statistics, and robustness. These questions are also crucial for analyzing inference in the presence of identification problems.(C) Causality determination and measurement constitute important issues for prediction and policy analysis in economic time series. I wish to pursue my earlier work on these questions in two directions: (1) by allowing for nonlinear restrictions entailed by economic theory (which may also raise identification difficulties); (2) by developing methods that allow for heavy tails (big shocks),
Programme
Discovery Grants Program - Individual
Bénéficiaire
Emplacement
Quebec, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data
Tel que publié
Application id
8581-2011
Name nom
Dufour, JeanMarie
Department département
Economics
Organization id
61
Institution établissement
McGill University
Province en
Québec
Province fr
Québec
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2016
Competition year année de concours
2011
Award amount
28000
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
1508
Committee name en
Mathematics and Statistics
Committee name fr
Mathématiques et statistique
Area of application code
1212
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Social sciences
Area of application fr
Sciences sociales
Research subject code
3000
Research subject group en
Statistics and probability
Research subject group fr
Statistique et probabilité
Research subject en
Statistics and probability
Research subject fr
Statistique et probabilité
Partie
5
Nb partie
5
Application title
Statistical inference with parameter indeterminacies, robustness and causality
Keyword
NSERC
Application summary
In this program, I intend to pursue research in three wide related areas:(A) statistical inference in the presence of identification problems;(B) robustness to model incompleteness(C) causality analysis in time series(A) In the first area, I wish to pursue my earlier work on the statistical problems associated withobservationally equivalent parameter vectors (identification) which often occur in many statistical models, for example structural models derived from economic theory which suggest that explanatory variables in a regression should be viewed as ""endogenous"". Even though it is commonly accepted that standard inference procedures (estimators, tests, confidence sets) become invalid in such models, I wish to take a contrarian view and explore in a systematic way when and in what ways standard inference procedures not designed for structural models [such as ordinary least squares estimators, maximum likelihood, various testing procedures, etc] remain ""valid"" in such models (if not possibly superior to methods designed to deal with such models) - with or without ""strong identification"" - to what extent forecasts based on such methods can still yield useful results, etc. This research may require redefining somewhat what is meant by a valid inference. Related problems that will be studied include:(1) the relationship with set identification and testability;(2) instrument selection, taking into account the fact that methods specifically designed for inference on structural models (such as IV methods) may be inferior in finite samples to methods that do not take ""endogeneity"" into account; (3) applications to factor analysis; (4) applications to inference on dynamic general equilibrium models.(B) In the second area, I wish to study the statistical theory associated with inference on incompletely specified models: what restrictions are needed to ensure that a statistical procedure remains valid despite the fact that the process governing some variables is not specified (or is incompletely specified)? This issue will be explored by using methods based on the theory of invariant statistical procedures, sufficient statistics, and robustness. These questions are also crucial for analyzing inference in the presence of identification problems.(C) Causality determination and measurement constitute important issues for prediction and policy analysis in economic time series. I wish to pursue my earlier work on these questions in two directions: (1) by allowing for nonlinear restrictions entailed by economic theory (which may also raise identification difficulties); (2) by developing methods that allow for heavy tails (big shocks),