1.1 PURPOSE OF THE HANDBOOK
The purpose of the handbook is to serve as a reference guide and to provide a broad overview of the statistical processes. The audience is beginners and others who want to undertake a new statistical data collection or those who wish to improve their understanding of statistical processes involved in such collections.
The handbook provides an overview of key issues to be considered for:
· conducting statistical surveys;
· extracting data from administrative collections;
· managing statistical data;
· transforming administrative or survey data into statistics;
· analysing and interpreting statistical data; and
· disseminating results
The information presented is in accordance with the NSS Key Principles which articulate 'best practice' guidelines in order to promote high standards in the collection, compilation and dissemination of statistics. The ultimate objective is to assist the commonwealth, state/territory and local governments and government organisations in their statistical activities.
1.2 SURVEYS AND ADMINISTRATIVE COLLECTIONS
This handbook covers both statistical surveys and administrative data collections.
Statistical surveys collect data to answer a question by placing a statistical value to an indicator or measure. Surveys measure one or more characteristics of the whole population or a sample of the population.
Administrative data collections are primarily held by organisations for internal use, for example, school enrolment, hospital separation, birth, death and marriage records. The range of information collected and held by organisations is determined by their requirement to manage their work programs.
Although the primary purpose of administrative data is for use in the administration of work programs, this does not preclude the use of such data for other statistical purposes. In fact, administrative collections are a rich source of data for many statistical analytical purposes.
Administrative data can however be difficult to interpret as statistical information. This problem can be addressed by enhancing existing administrative collections. Many of the principles governing good practice in statistical surveys are also applicable to administrative collections.
The structure of the handbook uses a statistical cycle model comprising seven phases of statistical activity including planning, developing, collecting, processing, transforming, disseminating and evaluating the process.
The handbook also provides important information on issues such as ways to handle confidential data, use of standards and classifications and managing data quality which underpin the statistical collection process.
The seven phases of the statistical cycle discussed in detail in subsequent chapters in the handbook are:
In this phase decisions should be made regarding the objectives of the collection, budget and resources, timeframe, training, confidentiality etc. Research into the subject matter, statistical process and existing data is generally conducted prior to undertaking a new statistical collection. This phase also involves determining legislative requirements and establishing a project management plan.
Statistical collection development is concerned with developing the survey frame and sample design and methodology. It also involves determining the most appropriate data collection method, consulting with stakeholders to determine their needs and designing any questionnaire. Survey testing at this stage can provide guidance on many aspects of the survey design and content. Opportunities to reduce respondent load can be considered in this phase. Making data specifications and identifying output requirements are also part of developing the collection.
Data collection involves dispatch and collection control systems, maintaining collection documentation, providing interviewer training and instructions for face-to-face or telephone based data collections, monitoring response rates and the quality of the data collection process.
Data processing aims to produce datasets which are free from errors and that can be manipulated or transformed into statistics. This generally involves coding, checking, editing, and where necessary weighting. Privacy and confidentiality of data being processed should be also monitored during the processing stage.
Statistical analysis can be undertaken once estimates have been produced through data processing. This analysis can range from simple summary tables to more complex analysis such as time series analysis. Analysis requirements should have been ideally factored into the planning process.
Dissemination involves determining the most appropriate methods of delivery of the statistics or results of statistical analysis to users. Statistical information may be disseminated electronically or through conventional publications. The data needs to be confidentialised if necessary prior to release to ensure that confidential information about a survey respondent or business is not made public.
This phase involves evaluating the quality of outputs and devising strategies to improve the data quality for future purposes or to inform users of any quality shortfalls. It is also necessary to monitor and evaluate processes (e.g. editing, coding) and systems (e.g. business register, classifications) to determine the causes for quality problems to take remedial actions. Evaluation can involve assessment and analysis of both quantitative (e.g. survey response rates) and qualitative (post enumeration surveys) approaches.