Questionnaires can be a relatively inexpensive way to gather data from, potentially, a large number of respondents. Questionnaire development is a multi-stage process; a blend of scientific and artistic elements. Some of the issues that will be discussed are: (1) Identifying the objectives of the survey (Are you trying to discriminate, predict or evaluate? Census or sample or secondary data?) (2) Determining the method of administration (In-person/self-administered/by phone/web-based/mixed methods) (3) Writing the questionnaire/cover letter/script (Open or closed questions? Phrasing, layout and flow; Pre-testing) (4) Collating and interpreting the findings.
Social network data are unique because of their focus on relationships between people instead of attributes of single individuals. The presentation will show: ● a general overview of the social network approach ● the uniqueness of collecting and analyzing social network data ● how social network data can be re-coded and imported into Stata for further analysis, and ● an example to illustrate the procedure. No prior knowledge of social network analysis is required as this presentation only covers basic concepts and procedures.
Missing data is a common problem in quantitative social research, particularly with large multivariate models and/or sensitive topics. Deleting cases with missing data is often problematic and can lead to smaller and biased samples. We will discuss the merits and limitations of different approaches to deal with missing data in multivariate studies, including traditional methods (e.g. single imputation, dummy variables), modern methods (e.g. multiple imputations, Expectation-Maximization), and treating missing cases as a substantive dependent variable of interest.
The presentation will discuss the use of simulation methods in empirical work in the social sciences. The methods will involve computer simulations that can be implemented with commonly available software, e.g. Excel, STATA, SAS, MATLAB, and will be useful in estimating statistical models, understanding variability in model estimates, and drawing inferences from observed data.
There are many ways to explore data through graphical displays. R, the open source statistical package, supports many of them: static graphics such as histograms and scatterplots and multi-panel combinations thereof, as well as interactive graphics. This presentation will give a tour of the capabilities.
Building on the presentation on October 26 regarding the access to Public Use Files, this presentation will demonstrate how users can gain access to Statistics Canada’s Microdata Files. These files are non-anonymized, and sensitive data such as actual income are not suppressed or categorized. Longitudinal data like the National Longitudinal Survey of Children and Youth (NLSCY) are also available for analysis of trends across time. In Canada, these data are currently accessible through the Research Data Centers in university campuses across the country. The presentation will highlight some of the datasets and explain how researcher can gain access to these data.
Western Libraries Map and Data Centre: Accessing Data at Western using Equinox, <odesi>, and Statistics Canada's website
Vince Gray, MLIS & Elizabeth Hill, MLIS
Western faculty and students have access to a wide array of data for research and teaching purposes through agreements such as the Data Liberation Initiative and the Inter-University Consortium for Political and Social Research. This presentation will show how to locate and download data from three sites: (1) Western's Equinox Data Delivery System, (2) Ontario Council of University Libraries' <odesi> database, and (3) the web site of Statistics Canada. The presentation will also describe differences in products offered at the sites, and demonstrate access to both microdata and aggregated data products such as Beyond 20/20 files from the 2006 Census.
The presentation will introduce basic time-series models, such as the Auto-Regressive Moving Average (ARMA) model and the Generalized Auto-Regressive Conditional Heteroscedasticity (GARCH) model. It will focus on some of the issues that arise when working with time-series data, and include a brief demonstration of the popular software for time-series analysis.