Each of the specific hypotheses, except for those that referred tothe FED architecture,15 was tested using separate binary logisticregression analyses. The results of the hypotheses testing arepresented in Table 5. As illustrated by a strong support for H1, H4,H5A, H6A, and H7A, when considering the choice between IDM andEDW architectures, information interdependence, strategic view,resource constraints, perceived ability of the in-house IT staff, andlevel of sponsorship each influence architecture selection. Severalhypotheses were not supported: H2A, H2B, H5B, H6B, and H7B. Asmultiple tests were being conducted on the same sample, in orderto control for family-wise and comparison-wise errors the p-valueswere adjusted for multiple testing using the Benjamini et al. [19]procedure. As the adjusted significance levels for the hypothesestests indicate (see Table 5), the previous hypotheses testing resultswere significant except for H7A. As the initial hypothesis testresults indicate, H7A was marginally significant at a p-value of0.049. It is possible that low sample size may have affected thetesting of the influence of sponsorship on the selection betweenIDM and EDW.