For operational modal analysis (OMA) the multi-setup (or roving sensors) measurements strategy are often applied because the number of wanted measurement DOFs exceeds the number of available sensors. Studies of potential effects of applying the multi-setup strategy in contrast to the optimal simultaneous measurement strategy have, however rarely been reported. Classically, multi-setup datasets are merged together in a post-identification step, known as the patch based approach in experimental modal analysis. Recently, methods for merging the datasets in a pre-identification step has been suggested in the literature, but none of these methods have been compared with simultaneous measurement data. The present paper presents experimental results from an OMA obtained from a measurement consisting of 45 sensors simultaneously measured on a ship in operation. The modal parameters have been estimated for the full dataset, and for three artificially created multi-setup tests based on the full dataset. The first multi-setup test is composed of five datasets that are parallel in time, hence consistency between the datasets is guaranteed. The second test is also parallel in time but with two datasets, and the third test is composed of the same two datasets, but serial in time, hence consistency between the datasets is not guaranteed. For the multi-setup tests merging of the datasets have been done both with pre- and a post-identification methods and the modal parameters are compared with the ones estimated from the simultaneous measurement data. It is shown that using multi-setup tests systematic errors can occur especially in the damping estimates. The tested case where consistency between the datasets could not be guaranteed larger error was observed than when consistency was ensured. In addition, it was found that when mode shapes are of high importance the number of measured DOFs in each dataset should be sufficient larger in order to have a good estimate of the mode shapes.