Specify and generating survey questionnaires using active metadata

Keywords: survey, generate web questionnaires, metadata, meta-driven approach, DDI

1. INTRODUCTION

Insee has been developping during the last years a web based survey platform for the data collection with a meta-driven approach. The first step is the possibility of responding online via a single web portal : the Coltrane Project (acronym for COLlecte TRANsversale d’Enquêtes » in French, Cross-Disciplinary Survey Data Collection)1. Coltrane is a data collection platform, currently use for web establishment surveys. It is composed of several modules including Eno, a tool generating web questionnaires from their formal description, conformant to the DDI standard2.

A recent enhancement to this generator is a questionnaire designer for survey conceptor, named Pogues. More specifically, Pogues produces the DDI description of the questionnaire that is then submitted to an embedded instance of Eno. Using this tool, a survey manager or a survey designer can specify his questionnaire in a friendly way and visualize the generated result immediately. Only the web questionnaire is available now, but different formats will be offered in the future..

In parallel, a global metadata repository, named Rmés, has been set up to manage metadata across the entire life cycle of a survey, from design to dissemination and further uses. It specifies all relevant metadata, ready for use later during all the statistical processes, as well as quality assurance procedures.

Ongoing work focuses to link Pogues with the global metadata repository to stock, reuse and share metadata and harmonize the process.

2. METHODS

As the Coltrane project bagan a few years ago, the main challenges were :

- to extend, modernise, industrialize and standardise online data collection for all establishment surveys ;

- to set up a metadata-driven approach. It reduces development costs, permits to have the documentation of the survey always in sync with the questionnaire in production. And beyond building collection tool, this principle could be implemented on other sub-processes in the GSBPM ;

- to provide a single access point for establishment and the ansewer easier ;- to enable establishment to better identify surveys that are actually within the

official statistics remit.

1 A. Husseini-Skalitz proposed a abstract about COLTRANE for the NTTS 2017 whose title is “Modernisation and industrialisation of the online data collection”

2 http://www.ddialliance.org/1

The first release of Eno aimed at creating different types of questionnaires for the Structural Business Statistics survey. The innovation part of this operation was to produce the form in Open Document Format (then in fillable PDF) from its metadata description in DDI 3.1, using a generator. The main goal was to prove that this way of generating questionnaires from their metadata description was possible and effective. The whole chain was acchevied in February 2013.

The second release of Eno was able to generate web questionnaires from their description in DDI 3.2 ; those questionnaires were online in March 2015 for another business survey. Many other surveys have already been integrated and some others will be integrated over the next years.

The Eno service is developed using XSL transformation to build a Xforms file for the Coltrane portal (implements the form runner Orbeon). It is open source, internationalized, and now available as a CSPA3 service. Future work will allow to perform significantly more surveys by providing new output formats for Eno (paper very soon, Blaise) and a bigger functional coverage for specifics household surveys.

The choice of DDI as a modelling format for questionnaire specification was motivated by :

- A strong need for a precise modelisation of the questionnaire in order to be able to render its content and logical flow in different collection modes (i.e a semantically rich model);

- DDI being an open standard;

- DDI being serialised in XML, and thus easily processable.

This type of questionnaire would also allow us to incorporate dynamic controls to check the format of the input fields, consistency between the different input fields and, where necessary, with previous data.

Pogues produces the DDI description of the questionnaire that is then submitted to an embedded instance of ENO. Pogues is developped according to an agile method with frequent releases. The current version supports the main functionnalities needed for complex business surveys, and the roadmap foresees the developpement of more complex logic flows that can for example be found in household surveys. Other future developments will enhance the possibilities to specify questionnaire controls that could be used not only during the data collection process but also during the data editing and imputation processes.

In parallel, ongoing works focuses to link Pogues with the global metadata repository Rmés to stock, to reuse metadata a year on year and to share questions, code lists,... between questionnaires. Registering metadata in an early step of a survey allows to propagate metadata on other sub-processes in the GSBPM.

Pogues is developed using React/Flux framework and additional JavaScript tools. It is open source and natively internationalized, and can be therefore used directly by other statistical agencies. Several national statistical institutes have already shown their interest in this software.

3 http://www1.unece.org/stat/platform/pages/viewpage.action?pageId=1121328352

These tools are presented in the ESSNet “Shared services for European statistics”4.

3. RESULTS

Specify your questionnaire with Pogues :

Questionnaire generated by Eno :

The demonstration of the full functionnalities of Pogues and Eno requires an access to the Insee network, and thus a secure laptop will be throught by the prensenter. Only a regular Internet connection will be necessary.

4 http://ec.europa.eu/eurostat/web/ess/about-us/ess-vision-2020/implementation-portfolio#SERV3

4. CONCLUSIONS

Currently, nearly 200 000 questionnaires are produced by Eno each year, and this figure is expected to grow to more than one million questionnaires, corresponding to 34 surveys, within the next two years. Ongoing work focuses to wide functional coverage for specific household surveys and link with the global metadata repository Rmés.

A full meta-driven approach allow to have a standardized approach to statistics production process while increasing data quality and reducing workload.

REFERENCES

[1] A. Husseini-Skalitz, “Modernisation and industrialisation of the online data collection”, proposed a abstract for the NTTS 2017 

[2] T. Dubois, “Pogues & Eno generator : Specify and view a web questionnaire generated from its formal description”, International Conference on Establishment Surveys (2016).

[3] É. Sigaud, R. Tailhurat, F. Cotton and É. van der Vlist. “XForms generation: a real world example.” Presented at International Symposium on Native XML User Interfaces, Montréal, Canada, (August 5, 2013). In Proceedings of the International Symposium on Native XML User Interfaces. Balisage Series on Markup Technologies, vol. 11 (2013).

4