See Fig. 1.

Figure 1.  

Tobacco Data in Africa Project Data Inventory 2016. Original data available as Suppl. material 1.

Data Collection methods

1. Desk-based search of official websites of project countries

This will involve searching websites of government departments These will include data on: 

1. Tobacco production (Departments of Agriculture)
2. Prevalence of tobacco-related diseases, and tobacco-related morbidity and mortality (Departments of Health)
3. Tobacco taxation (Internal Revenue Services)
4. Tobacco products manufacturing, tobacco imports/exports (Departments of Trade and Industry)
5. Tobacco usage, from Surveys by National Statistics Agencies (NSAs). In South Africa unit record administrative data from government departments, repackaged as research datasets, are also shared by the NSA. If data collection instruments (administrative forms) used to collect the data are available on these sites they can provide useful information on the data.

2. Desk-based search of websites of International Development Organisations

This desk-based study will allow us to discover tobacco data on project countries that has been collected by international organisations. The will include the websites of the international DHS Program and UN bodies such as the World Health Organisation (WHO). From this a "question bank" will be created of useful variables and the datasets where these can be found.

2. Desk-based search of industry websites

The third component of our desk-based research will involve examing online records of the tobacco industry. From these we hope to obtain data on: Cost of tobacco production, and profits, in the industry, prices of raw tobacco and tobacco products, salaries, capital and foreign investment, mergers and acquisitions, advertising spend, and regulations in the industry

3. Approaches to data holders

We will create metadata on our platform for surveys already shared by others online. Our desk search may also reveal the existence of datasets with a tobacco data components but which are not in the public domain. In these cases we will approach the relevant research projects in the project countries to release this data and allow the Project to host this on their Open Data portal. This may be a fraught process but any challenges and successes can be written up to inform our future work.

4. Own surveys

The project has already crowd-sourced data on current prices of tobacco products in two project countries. This may be expanded during the course of the project to all project countries. We will upload and share the metadata and data from these surveys.

Supporting documents

Data collection and data analysis documents will be shared along with the data files, where available. Forms used for collecting administrative data will be shared with administrative datasets. Data collection instruments (questionnaires, diaries) will be made available with the survey data. Code lists used in collecting the data will also be provided. Final reports from data collection projects will also be shared, where available.

Metadata

Each dataset will have a metadata record to help data users analyse the data. This metadata record will be created during examination of the data and data collection instruments. It will include information gathered on the dataset during the data collection process. This documentation can be an invaluable source of provenance and usage information for those analysing the data. Notes on data quality will form part of the metadata record. Metadata will be created according to the Data Documentation Initiative (DDI) international metadata standard, using Nesstar Publisher, which is free data markup software for the creation of XML-compliant metadata according to the DDI standard.

The administrative data we will collect will mostly be in the public domain, in the form of reports and other records from government departments. The survey data we will collect will be anonymised data already shared with researchers, although not always online. The industry data will be data made available to shareholders and the public. We are adding value by bringing these sources together and providing a means for researchers to easily discover and download these data.

However, we will endeavour to make data available that is not yet in the public domain. In these cases, we will ensure that:

1. We have the necessary permissions from data owners to make these data open.
2. The data is suitably anonymised, to protect respondent confidentiality and privacy
3. We take national laws on sharing data across borders into account. Where such restrictions exist, we will be unable to host this data.
4. We work with all stakeholders to ensure agreement on what will be shared, how, and with whom.

The government data we will collect is not subject to IPR. The tobacco industry data we will collect will be public records. We will therefore not publish information that would compromise any IP rights. However, we will check each traunch of data we obtain, to ensure we have permission to pass the data on to third parties.

The data would be stored on a server managed and backed up by the University of Cape Town's Commerce IT Department. Curation of the data will be the responsibility of DataFirst's Research Data Service. DataFirst is a technical partner on the Project. Each preservation dataset will consist of data files, document files, metadata files, and any programme files used in creating the data files. Data Service staff will be responsible for adding data updates to datasets. We will also handle version control to ensure the most recent and accurate data files are published, and provide tombstone citations to earlier versions for verification or replication of research which may cite these supercede versions of the data.

Access to the server hosting the preservation datasets will be password controlled. Passwords will be allocated by the Commerce IT manager only to Data Service staff. Server software will monitor data security and integrity.

Criteria for preservation will be:

1. Data is tobacco-related
2. Data covers project countries
3. Data is accurate and reliable (we will undertake quality audits to determine this)
4. Data is unit record data (not aggregated but available at the level at which it was collected)
5. Data is not readily available from another repository

Retention:

It is difficult to predict what data has long-term value. Our policy will be to store unit record tobacco data indefinitely. As these datasets grow, so will their value over time. Time-series data will continue to be useful for economic and health policy research in the long term.

Sharing:

Because we aim to establish an Open Data portal, all data retained/preserved will also be shared. The Project's policy is aligned to DataFirst's policy: We do not archive data which cannot be shared with researchers in some form and at some access level.

There will be numerous datasets. Our long term preservation plan for the Project's data holdings depends on the sustainability of DataFirst's Research Data Service. The service was established in 2001 and is a unit at the University of Cape Town, a well- funded and well-established university in South Africa. Our sustainability prospects are therefore good.

The data will be shared as discrete datasets (by country, year, data source). DataFirst hosts and shares data via an online dissemination platform, based on the National Data Archive Open Source software developed by the World Bank's Development Data Group.

The platform provides a number of data access options. The Project's data will be shared as Public Use data. That is, researchers will need to register on our site and say for what purpose they will use the data, but access will be immediate and automatic, with no vetting of use. The usage information we collect will be to support service improvements.

The data will be shared in a number of formats, includind excel spreadsheets, and data files in the commonly used statistical analysis programmes (SPSS, Stata). We will also make the data available as .csv files, in line with Open Data requirements.

We aim to share the tobacco data we collect as public access data. We do not aim to support research-use only requirements, as this is counter to Open Data principles. Policy research, academic research, business analysis, and private sector innovation all need good data, and countries benefit from informed decision-making in all these spheres.

The Manager of DataFirst will be responsible for curating the Project's data. This is in line with the project proposal for DataFirst to be funded to provide technical support. DataFirst's Manager has 25 years' experience in managing research data and working with data users. Skills learned from undertaking data rescue projects in South Africa will also be useful in assisting with data collection activities.

Funding for data collection as been budgeted for in the Project. This may need to include funding to travel to project countries and negotiate with data collectors in government and academia to release their data, and allow its reuse. Funding has been provided for a Project Manager. The Project Manager is responsible for conducting online data audits and downloading data, and populating the database which DataFirst's Manager will curate. This will be a time- and labour- intensive task and more staff hours may need funded for this.