Survey Design B explores these topics:

• Designing the research questions

• Evaluating the chosen research questions

• Processing the data prior to statistical analysis

• Accounting for missing data and non-response

Session 1: Designing the research questions

• There are many obstacles for a participant when choosing their response to a survey question. These obstacles include flaws in understanding what the research question is trying to ask, retrieving possible responses from their short-term and long-term memory, evaluating which pieces of remembered information to base their response on, and communicating their response to the survey interviewer. In this session we explore the difficulties that a participant will have in providing a response, and will describe how to develop a questionnaire to maximise the accuracy of provided responses.

Session 2: Evaluating the chosen research questions

• Following the design of a survey questionnaire, the questionnaire is then presented to a pilot audience in order to identify potential weaknesses in the survey. This piloting can include presenting and discussing the questionnaire with application experts of members of the target population, or administering trial survey interviews with survey assistants or study participants. In this session we will explore these methods for evaluating the chosen research questions, and describe how to maximise learning through this evaluation process.

Session 3: Processing the data prior to statistical analysis

• Between data collection and statistical analysis there are a number of tasks that must be performed with the survey data. These include coding a non-numeric answer (such as a paragraph of text) into numeric data (eg. categories), verifying that recorded data is consistent with expected responses (eg. checking that responses were provided in the same units of measurement), and weighting the data to account for decisions made in the sampling design. In this session we will explore the various tasks involved in data processing, and describe how to maintain quality of our dataset throughout this process.

Session 4: Accounting for missing data and non-response

• Missing data can occur for a variety of reasons – the inability to contact particular people that we would like to include in our survey, potential participants refusing to participate in the survey, and participants willing to participate in the survey but who are unable to provide the requested information. There are numerous approaches for addressing concerns over missing data that can be employed during the design of the survey and the questionnaire, during the interaction between the interviewer and the participant, and during the data analysis. In this session we will explore the various approaches for accounting for missing data, and review their strengths and weaknesses.

These workshops will use a combination of three teaching styles:

• Lectures

• Group discussions

During the lecture sessions the theory of survey design will be presented, and will be discussed in an interactive manner with the class.

During the group discussions the class will be involved in both designing a survey and also in being the participants within the survey. Through such a process a number of practical issues involved in survey design will be explored.

This workshop is intended for audience members who are new to the methods of survey research, and who desire a solid foundation in these methods. A simple knowledge of statistics is assumed (where audience members should be comfortable discussing terms such as mean and variance).

These workshops are designed for a broad range of participants, researchers from various application fields who only spend a small portion of their time doing statistics (but researchers who when they do statistics want to be doing statistics well). Communicating complex statistical ideas to a non-statistical audience requires significant skill. Equations for example are vital for statisticians who want to describe statistical models accurately but succinctly, however these equations can be a stumbling block for a larger audience who are not used to working with complex equations. We think one sign of a good statistician is that they know the pictures that go on behind the equations, and in our workshops we focus on the pictures rather than the equations.

Working with this broad audience we teach complex statistical theory (we find it heart-breaking when excellent research studies are undone by a poor understanding of underlying statistical theory), but we teach this theory using a number of practical real-world examples. We are also a strong believer in the use of humour in a training context, where humour means that participants are more likely to relax in a training context, to engage with the training content and the presenter, and are more comfortable to ask the questions that they really want to ask about the content as a result.

Each workshop will be capped at 20 participants.