Events and activities
Network activities, events and meetings
The kick-start meeting was held at the United States Census Bureau in Washington DC (26-27 February 2015).
At this meeting, each partner provided a presentation on their institutional state-of-the art. In addition, a research agenda for the network was widely discussed and was finalized in April 2015. The research agenda document is in Appendix 1 of the BADEN 2015 annual report.
A bi-lateral meeting between CBS Netherlands, University of Manchester and Statistics Sweden was held in the Netherlands (28 - 29 May 2015).
At this meeting, theoretical topics around the Bayesian framework for adaptive survey designs were widely discussed and clarified. The framework is decomposed into three separate components: modelling design parameters (response propensities, cost functions); learning and updating posterior distributions of design parameters based on information coming in from the current survey; optimisation of quality and cost criteria that also have probability distributions In addition, the different types of adaptive survey design strategies handled by partner institutions were classified: (1) predetermined strategies (eg. recruitment letters, mixing data collection modes); (2) follow-up strategies (eg. interventions during data collection, stopping criteria); (3) hybrid of both strategies. Other discussions related to partnerships and PhD students, joint research and possible sharing of data. The first theoretical paper is led by researchers at CBS Netherlands.
The Intermediary meeting with all partners was held at Statistics Sweden in Stockholm (10 – 11 September 2015)
At the meeting, the US Census Bureau partner presented an overview of a study comparing the many indicators presented in the literature that have been used for adaptive survey designs. Initial results show that many of these indicators measure the same thing, eg variation in response propensities, which is indicative of a lack of representativeness. Each of the partners then presented their institutional case study for the elaboration of the Bayesian framework for adaptive survey designs. These case studies will form the basis of the work within the network. Discussions were held about collaborations, possibility to share data and proposed papers for the first network dissemination at the 71st Annual American Association of Public Opinion Research (AAPOR) Conference in May. CBS Netherlands partners provided an overview of the first theoretical paper on Bayesian Adaptive Survey Designs which was widely discussed. Finally, all partners discussed planning for the 4th International Workshop: Advances in Adaptive and Responsive Survey Design to be held at The University of Manchester, (9 - 10 November 2015).
October, 2015: The PI successfully proposed an invited panel session to the 71st Annual American Association of Public Opinion Research (AAPOR) conference, Austin Texas, 12-15 May 2016.
4th International Workshop
Advances in Adaptive and Responsive Survey Design at The University of Manchester (9 - 10 November 2015)
There were 40 participants to the international workshop from the following countries: United States, Italy, the Netherlands, United Kingdom, Germany, Canada, Norway, Sweden and Australia. The Leverhulme International Network grant funded two invited speakers: Professor Rod Little from the University of Michigan and Professor Carl Sarndal from Stockholm University. Additional funds were obtained through the ESRC National Centre for Research Methods to fund Professor Roger Tourangeau from Westat, United States. In addition, Professor Peter Lynn from the United Kingdom was also funded form internal funding. There is no doubt that the participation of these leading and prominent professors greatly enhanced the quality of the discussions and feedback on the papers that were presented at the workshop. The agenda and presentations can be found here.
A bi-lateral meeting between the University of Manchester, Statistics Netherlands, University of Michigan and the US Census Bureau, was held at the US Census Bureau in Washington DC (21 – 22 March 2016).
At this meeting, each country reported on their case studies and discussed the presentations for the invited session at the 71st Annual AAPOR Conference, Austin Texas in May 2016. We also discussed the first theoretical network paper led by Statistics Netherlands on the general Bayesian framework paper titled: A Bayesian Analysis of Design Parameters in Survey Data Collection. We heard from the University of Michigan on their research related to response propensity modelling for daily survey response monitoring under a Bayesian framework. Whilst visiting at the US Census Bureau, we also heard talks from members of staff on adaptive design applications for both the Survey of Income and Program Participation (SIPP) and for the US Decennial Census. It was very interesting to hear about some of the problems encountered when transforming research to practice.
American Association for Public Opinion Research 71st Annual Conference (AAPOR 2016)
The intermediary network meeting and invited session were held during the 71st Annual AAPOR Conference, Austin Texas (12 – 15 May 2016).
The invited session was held on 15 May 2016. The session was well-attended and a great success with positive feedback from conference participants. The presentations from the invited session are available here.
An additional bi-lateral meeting between Statistics Netherlands, University of Southampton, Statistics Sweden and the University of Manchester was held at Statistics Netherlands (20 - 21 September 2016).
At the meeting, the Statistics Netherlands team demonstrated the software application in ‘R’ for the Bayesian estimation of design parameters which will be used for each case study, starting with the implementation on the case study of Statistics Sweden based on the Swedish Labour Force Survey. We also obtained updates on the first theoretical network paper from Statistics Netherlands. We held extensive discussions on the case study from the University of Southampton based on the longitudinal dataset ‘Understanding Society’, where Bayesian modelling is applied for estimating response propensities using information from previous waves.
Intermediary Network meeting
The intermediary network meeting was held at the University of Southampton (14-15 November 2016).
At the meeting we obtained updates on country case studies and discussed some initial results and findings from the first stage of estimating design parameters under the Bayesian framework: how to identify and weight the prior distributions for estimating the design parameters and the influence of the prior distributions on the posterior distributions. We are moving to the next stage of research which is on the optimization and prioritization of adaptive survey designs under the Bayesian framework. A draft outline of the second network paper was written by Statistics Netherlands and distributed to the network partners. We also discussed the planning of the 5th International Workshop on Advances in Adaptive and Responsive Survey (ASD) to be held at the University of Michigan on November 6-7, 2017 and the development of software tools in ‘R’ with a manual as one of the key outcomes of the project
2-3 March 2017, Washington DC
Joint Statistical Meeting (JSM 2017), invited paper sessions to disseminate research findings
26 July – 3 August 2017, Baltimore
5th International Workshop on Adaptive and Responsive Survey Design and BADEN summary meeting
6-7 November 2017, Michigan