Gönüllü web anketlerinde yanlılığı azaltmak için ağırlıklı düzeltme tekniklerinin karşılaştırılması

Web surveys have become one of the most widely utilized and popular survey methods in recent years. A web survey is performed over the World Wide Web by inviting individuals to complete the questionnaire by themselves. Internet usage has increased among inhabitants of developed countries as well as developing countries. Since the advent of the smartphone, web surveys have become even more viable and reasonable data collection method. Although, in the last years, internet-based data had been collected only for marketing researches. Data collection on the internet is faster than other methods such as paper-and-pencil, computer-assisted telephone interviews and personal interviews, because it is simple, cheap and provides quick access to the desired large group of respondents. However, in web surveys, bias may arise mainly due to limited coverage and self-selection. This study appraises characteristics of web surveys, their importance and trend, as well as problems to identify their bias. Some weighting adjustment techniques for reducing the bias are illustrated, and non-probability estimates from volunteer panel web surveys are compared to random sample estimates. Post-stratification weighting, generalized regression modeling, raking ratio estimation and propensity score adjustment techniques are used for reducing these biases. In the application of this study, population estimates are compared to those from a volunteer panel web survey. The data are from the survey "Using Social Networking Sites in the Education of Students of Open Education System, Anadolu University". A random sample based on simulation is created by a stratified probability sampling design, and estimates based on this sample are compared with estimates from a non-probability-based volunteer panel web sample. It is shown that the weighting adjustment has reduced the bias substantially and the random sample estimates provide better results than those from the volunteer panel web survey. However, when it is necessary to use volunteer panel web surveys, it is recommended to adjust at least one of the weighting adjustment techniques described in this study. Keywords: Web surveys; Volunteer panel web surveys; Probability and non-probability sampling design; Random sample; Bias; Weighting adjustment techniques.