Web Survey Bibliography
Whenever two or more survey statistics are compared, the question arises whether this comparison is warranted. Warranted usually means that there is no methodological artifact that could possibly explain any differences: I term this the “strong” interpretation of comparability. The “weak” interpretation of comparability is then that artifacts might exist, but evidence shows that they are not strong enough to explain away a particular substantive finding. In this chapter I discuss some methods to prevent, detect, and correct for incomparability. Translation issues and coding of design characteristics of questions in different countries are particularly relevant to cross-cultural studies. Strong and weak comparability, and the methods associated with them, are discussed for different aspects of total survey error (TSE). On the “measurement side” of TSE, invariance testing, differential item functioning, and anchoring vignettes are well-known techniques. On the “representation side,” I discuss the use of the R-indicator to provide evidence that the comparison of survey statistics is warranted.
Springer Homepage (abstract) / (full text)