Journal: Journal of Chemometrics, vol. 16, p. 81–88, 2002
International Standard Numbers:
Open Access: none
Data from ripening experiments of herring carried out at three Nordic fishery research institutions in the period 1992-1995 were collected and analyzed by multivariate analysis. The experiments were carried out at different times, with different stocks as raw material, using different types of treatments and analyzed in different laboratories. The question considered here is whether these data can be assumed to be one homogeneous set of data pertaining to ripening of salted herring or whether data from different labs, stocks, etc. must be considered independently. This is of importance for further research into ripening processes with these and similar data. It is shown in this paper that all data can be considered as one homogeneous data set. This is verified using resampling where latent structures are compared between different sample sets. This is done indirectly by testing regression models, that have been developed on one sample set, on other sample sets. It is also done directly by monitoring the deviation in latent structure observed between different sample sets. No formal statistical test is developed for whether samples can be assumed to stem from the same population. Although this can easily be envisioned, it was exactly the need for a more intuitive and visual test that prompted this work, developing different exploration tools that visually make it clear how well the data can be assumed to derive from the same population. Subsequently analyzing the data as one homogeneous group provides new information about factors that govern the ripening of salted herring and can be used in new strategic research as well as in industrial practice. Copyright (C) 2002 John Wiley Sons, Ltd.