ESSnet on Small Area Estimation Loredana Di Consiglio Fabrizio Solari
Istat
ESSnet Workshop, Stockholm 2010
Partners Istituto Nazionale di Statistica Institut National de la Statistique France et des Statistisches Bundesamt Centraal Bureau voor de Statistiek Statistisk Sentralbyrå Główny Urząd Statystyczny Instituto Nacional de Estadística de España Office for National Statistics Swiss Federal Statistical Office
(ISTAT) Etudes Economiques (INSEE) (DESTATIS) (CBS) (SSB) (GUS) (INE) (ONS) (FSO)
(will join the project) Stefano Falorsi, Istat, is the coordinator of the project contacts :
[email protected] [email protected] https://webgate.ec.europa.eu/fpfis/wikis/display/essmethodology/ESSnet+on+Small+Area+ Estimation
ESSnet Workshop, Stockholm 2010
Needs for Small Area Estimation Basic diagram:
U a population Ui a portion of the population, let us call it ith small area s a random sample from U The part of the sample s that falls in the small area Ui is si = Ui ∩ s The size ni of si is usually random and often small ESSnet Workshop, Stockholm 2010
When and How SAE? When to use SAE methods: Whenever direct estimator which is based only on sampling units observed for each small area are not reliable (small sample size or sometimes even no observed units) – i.e. variance or other measure of sampling variability considered to be too high for the indicator and level (area) under study
How to use SAE methods: Exploiting relationship among variables by means of explicit or implicit modeling
ESSnet Workshop, Stockholm 2010
SAE: Borrowing Strength from? How does SAE work? Borrowing Strength: Cross - sectional from Auxiliary data from Spatial relationship Over time
ESSnet Workshop, Stockholm 2010
SAE methods SAE is a collection of different methods: Synthetic methods (often implicit assumptions on the nature of relationship; a simple case: rates at NUTS2 level = rates at NUTS4 level)
Composite methods (linear combination of synthetic methods and direct estimators to balance bias and variance)
Estimator based on linear mixed models (EBLUP, EB, HB)
Non-linear models (e.g. logistic models for binary responses) with spatial and/or temporal correlation structures among random effects
Semi- parametric models
…
ESSnet Workshop, Stockholm 2010
Motivation for the project There is an increasing need for estimates at local and detailed geographic and socio-demographic level where direct estimation (based on sample data only) cannot ensure enough precision
nevertheless policies in Official Statistics with regard to the implementation and quality of SAE need to be defined Applications and experiences of SAE in Statistical Institutes overlap in terms of similar needs and sampling schemes, while may differ for external information Collaboration action creates synergies and allows everyone to benefit from the exchange of practices and know-how, representing the basis for identifying best practises in similar contexts ESSnet Workshop, Stockholm 2010
Motivation for the project
The project takes advantage from the results (in terms of methodology and routines for the implementation of estimators) already available from previous projects on small area estimation (e.g. EURAREA, SAMPLE, ...) The experiences of the ESS in small area estimation represent for all aspects of the project an important starting point for sharing knowledge The project aims at building a common reference framework for the development of SAE in the NSIs ESSnet Workshop, Stockholm 2010
Description of the Project The project is planned in phases which are a series of theoretical and application activities in order to facilitate and promote the use of small area techniques in the production of statistical information The results of the ESSnet project will be helpful to detect the best practises to define the guidelines to be followed for the applications of SAE to give some advices for software tools users. Dissemination of results is another important step of the ESSnet project, therefore a web-site will be available in order to share information and the results among the partners be a forum platform in order to boost the communication among NSIs willing to apply SAE methods.
ESSnet Workshop, Stockholm 2010
Description of the Project The project is composed by 7 work-packages:
WP1 - Project management
WP2
(GUS,Destatis, INSEE, INE, ONS) State of the
art The WP2 aims to provide a comprehensive overview of small area estimation in the ESS social surveys with respect to implementation, needs and expectations, the starting point will be represented by documents available on small area estimation, BUT the perspective of NSI will be highligthed; then the current application in UE NSIs and non-UE NSIs will be a central issue of the review
ESSnet Workshop, Stockholm 2010
Description of the Project
WP3 (INE, Istat, CBS, INE, GUS, ONS) Quality assessment
In order to support the choice and the use of SAE methods, this work-package aims to review and develop suitable criteria to assess the quality of SAE methods, in particular the study on methodologies for top-down assessment (from larger to smaller domains) and determination of accuracy thresholds, the review of methods for comparative assessment for the choice among different set of auxiliary information, the review on relevant methods for quality assessment (external validation using available information or MSE) and model diagnostic (model fitting, error analysis, etc.) As for WP2 the sharing of actual experiences of countries on the use of quality assessment would be the core of the WP
ESSnet Workshop, Stockholm 2010
Description of the Project WP4 – (Istat, CBS, GUS) Software tools this work-package aims at helping real application of SAE methods in NSIs, starting with a review of the available SAE routines and software on small area estimation, focusing on open source software (to make it available to all NSIs)
The WP will also provide recommendations for standardization and certification of ESS tools for SAE including a list of desirable aspects of the software tools. The study will be devoted to analyze the real capacity of routines to be applied to large scale surveys, i.e. characterized by a large number of individual records A product/deliverable of the WP will be the collection of existing routines and possibly the development of new routines
ESSnet Workshop, Stockholm 2010
Description of the Project WP5
(CBS- Destatis, Insee, Istat, INE, GUS,SSB) Case studies The activities of this WP will be focused on identifying a proper (sub-)set of SAE methods for each of the main survey framework (type of variable(s), sampling design). Relevant tools for diagnostics, identified in WP3, will be applied in real cases and WP5 will be the ground for training the software or routines produced in WP4 The project will be limited to social statistics applications Participating countries are mostly interested in LFS, but also victimization (CBS) and health statistics (Istat) will be considered in the applications WP6 – (Istat, CBS, GUS, ONS) Guidelines and recommendations The WP6 aims to summarise the activities and the results produced in the previous WPs so to provide practical guidelines in ESS social surveys context
ESSnet Workshop, Stockholm 2010
Description of the Project WP7 - (Istat- CBS -Destatis, Insee, Istat, INE, GUS, SSB,ONS) Transfer of knowledge and knowhow
The WP7 aims to transfer knowledge and know how to non participating NSIs and dissemination of results through a course and TOJ A web-site to update the information on the progress of the ESSnet to non participating countries (not yet available, on ESTAT platform?)
ESSnet Workshop, Stockholm 2010
Description of the Project Starting date: 22/12/2009- Duration in months: 24 months
ESSnet Workshop, Stockholm 2010
Communication between partners
wiki on Istat platform for sharing documentation and repository regular e-mailing between coordinator of the project and work package leaders meeting at the end of each semester to review work done and to plan work-load for the following semester Next meeting in Wiesbaden 5-6 July (end of first semester)
ESSnet Workshop, Stockholm 2010
Results of the first semester
WP2 – State of Art Review of literature completed, report to be finalised A questionnaire to explore the use of small area estimation
in NSIs
has been distributed (results expected the 7th of June)
WP3 – Quality Assessment Classification of the quality assessment (QA) methods completed First explorative search of relevant references on QA Questionnaire on use of small area estimation and quality
assessment to the ESSnet participants
WP4 – Software Questionnaire on use and practicability of SAE Software or routines submitted to ESSnet partners (deadline next meeting on 5-6 July in Wiesbaden) WP7 call for application for the training on the job has been disseminated
ESSnet Workshop, Stockholm 2010
Benefit of the ESSnet approach As stated above it allows synergies, exchange of practices and knowhow, representing the basis for identifying best practises in similar contexts It is particularly motivated by needs of NSIs, guided by the discussion carried out in DIME
Issues on SAE not exploited in the project but of great interest (to be developed in a future project?) this ESSnet will cover mainly cross-sectional methods; the temporal correlation in SAE methods may be investigated later. This issue is important for all repeated surveys such as LFS multivariate responses strategies for SAE at design and estimation level SAE for business surveys (now research activity in WP6 of BLUEETS, later the ESSnet structure after the end of the project may help the use in NSIs similarly as the current project)
ESSnet Workshop, Stockholm 2010
Helping success of the projects Communication among similar projects Borrowing strength from other projects on the same area, avoiding useless overlapping and exploiting common topics – exchange of results Inclusion of Academics as partners in the ESSnet (which is driven by NSIs needs) as a theoretical guide and internal “referee”
ESSnet Workshop, Stockholm 2010
