ACSPRI Conferences, ACSPRI Social Science Methodology Conference 2018

Font Size:  Small  Medium  Large

Bias Reduction for an Online Tracking Survey

Dina Neiger, Andrew C Ward, Tina Petroulias

Building: Holme Building
Room: Cullen Room
Date: 2018-12-13 01:30 PM – 03:00 PM
Last modified: 2018-09-21

Abstract


In 2016 the Cancer Council of Victoria wanted to investigate the implications of a change the survey mode for its tracking survey of smoking behaviour from a dual-frame RDD CATI survey to a nonprobability based online panel. This desire to change the sampling approach and mode of data collection was motivated by a range of factors including survey cost and time as well as the diminishing response rate of CATI surveys.
In this presentation, we report results from the study carried out by the Social Research Centre to investigate the accuracy of metrics derived from the nonprobability panel and to develop methods to adjust for the resultant biases using a four-step approach:
1. Identify a reference sample to be used to assess and calibrate nonprobability panel sample.
2. Develop pseudo design weights using the reference sample. Apply these to convert the nonprobability panel sample into a “quasi-probability sample”.
3. Identify key differentiators between nonprobability and probability sample and use these to calibrate the nonprobability sample.
4. Evaluate the bias reduction on key outcomes that were separate to the calibration process and measured in both the nonprobability panel and reference panel.