2016-Survey Methodology 2
- HYUKJUN (JAY) GWEON, University of Waterloo
New Approaches to Automated Occupation Coding Using Statistical Learning [PDF]
- Occupation coding refers to coding a respondent's text answer into one of hundreds occupation codes. Automated coding is a challenging problem because answers usually consist of only a few words while there are hundreds of possible categories. We develop several approaches that can be useful in the automatic coding context. The approaches include a hybrid method that combines a duplicate-based approach with a statistical learning algorithm, and a scoring method that is applied in a nearest neighbor approach. We illustrate the proposed approaches with occupational coding in the German ALLBUS panel survey based on the ISCO-88 standard.
- IBRAHIMA OUSMANE IDA, Laval University
Balanced Sampling by Using the Cube Method and the Rejective Algorithm [PDF]
- In recent years, balanced sampling techniques have experienced a renewed interest. They allow to reproduce the structure of the population in samples in order to improve the efficiency of survey estimates. New procedures have been proposed. These include the cube method, an exact method presented by Deville and Tillé (2004), and an approximate method, the Fuller (2009) rejective algorithm. After a brief presentation of these methods as part of an angler survey, we compare using Monte Carlo simulations, the survey designs produced by these two sampling algorithms.
- ISABELLE LEFEBVRE, Université de Montréal
Simplified Variance Estimation for Complex Designs [PDF]
- In a complex design framework, standard variance estimation methods entail substantial challenges. As we know, conventional variance estimators involve second order inclusion probabilities, which can be difficult to compute for some sampling designs. Based on Ohlsson's sequential Poisson sampling method (1998), we suggest a simplified estimator for which we only need first order inclusion probabilities. The idea is to approximate a survey strategy (which consists of a sampling design and an estimator) by an equivalent strategy for which Poisson sampling is used. We will discuss proportional to size sampling and two-stage sampling. Results of a simulation study will be presented.
- THUVA VANNIYASINGAM, McMaster University
Determining the Impact of Different Design Features on Relative Design Efficiency in Discrete Choice Experiments [PDF]
- Discrete choice experiments (DCEs) are used to quantify preferences of patients and health care providers. Guidance is needed to avoid creating designs with low statistical efficiency, resulting in biased preference surveys. We simulated 3204 DCE designs to assess how varying DCE design characteristics including the number of attributes (2-20), attribute-levels (2-5), alternatives (2-5), and choice tasks (2-20) affect relative design efficiency. Across all designs, more optimal designs were achieved with fewer attributes, fewer attribute levels, and more alternatives per choice task. These results are widely applicable for creating designs to elicit individual preferences on health services, programs, and products.
- WEI LIN, University of Toronto
Analysis of an Embedded Experiment in a Survey [PDF]
- We derive the Horvitz-Thompson estimator of the average treatment effect and its variance for a general design. In the presence of auxiliary information, a new model-assisted estimator for the average treatment effect is developed and the variance of the estimator is derived. We show that the new estimator is approximately design-unbiased when a general model is employed. Moreover, it doesn't require auxiliary variable information at the population level and is relatively easy to implement and compute. Simulations carried out indicate that the new estimator gains in efficiency and its relative bias is negligible.