Using equivalence testing to disentangle selection and measurement in mixed modes surveys
Understanding Society Working Paper Series
Mixed modes are becoming increasingly popular in surveys. This approach can decrease costs and non-response bias. But in order to evaluate the utility of this approach we must separate selection and measurement effects of the different modes. In this paper I propose a new way of applying the front-door method to control for measurement differences between modes: equivalence testing with latent measurement models. A small simulation study will show how this approach works and how it can be biased if the assumptions of exhaustiveness and isolation are not true in the observed data.