Evaluating relative mode effects in mixed-mode surveys: three methods to disentangle selection and measurement effects