Lukawetz, Gerhard

Improving Unit-Nonresponse Error Correction in Online Surveys by Using Multi-Dimensional Response Models


Abstract

A previous study (esp. Lukawetz 2002) has shown that factors like frequency of online use and response delay may be supportive in estimating direction and amount of nonresponse errors in online surveys. Since both the probability to contact potential participants successfully and their willingness to cooperate are closely related to individual online usage patterns, the two groups - "refusers" and "absentees" - are much more similar to each other in online surveys than they are in traditional face-to-face or telephone studies. However, distinguishing between these two effects may lead to a more in-depth picture of the response situation.

A two-dimensional response model leads to improvements in the identification and description of coverage losses. So it can be demonstrated, why late responding participants share more similarities with non-respondents than early responding. It does also make clear, why reminder messages fail to reach especially less frequent online users, thus enhancing coverage only of few cells of the model, leaving other cells still obscured.

This knowledge is used to formulate a nonresponse error correction model, trying to find the most similar substitutes for identified groups of nonrespondents. A survey described in Lukawetz 2002 provides extensive empirical data for a first application and testing of the model. Compared with two other approaches (based on item-nonresponse and structural-demographic weighting) this model proved to be by far the most effective technique for estimating and correcting nonresponse effects. However, including these approaches into a higher dimensional and therefore more complex model seems to be an interesting option.


Literature (Selection)
Cassel, C., Saerndal, C., Wretman, J. (1983). "Some uses of Statistical Models in Connection with the Nonresponse Problem." In: Madow, W., Olkin, I. (Eds.): Incomplete Data in Sample Surveys, S- 143-160, New York

Lukawetz, G. (2002): "Empirically Quantifying Unit-Nonresponse-Errors in Online-Surveys and Suggestions for Computational Correction-Methods." In: Batinic, B. e.a. (Eds.): Online Social Sciences, New York