In construction of estimators from survey data, one often encounters important issues arising from nonresponse. For establishment surveys, methods to address these issues generally must account for important features of the sample design and weighting structure. For any given nonresponse adjustment procedure, an analyst makes implicit or explicit use of models for the nonresponse phenomenon and the outcome variables of primary interest. The performance of the adjustment procedure then depends on the extent to which the data deviate from the assumed models, the impact of these deviations on estimator bias, and the inferential power of diagnostics designed to detect these deviations. This paper presents a simulation study to evaluate trade-offs among the issues of model deviations, estimator performance and detectability for establishment surveys.