In most of robust optimization models, uncertain parameters are assumed to be independent. For example Bertsimas and Sym or Ben-Tal and Nemirovski discussed that it is too conservative to assume that all of the uncertain parameters in a problem simultaneously take their worst values, and by this reason they introduced their famous uncertainty sets. However, if there is a kind of correlation between uncertain parameters, taking worst values by most of them will not be so unexpected. Furthermore if all parameters are completely correlated, we will expect that if one of them takes its worst value all other ones do the same. Therefore I think Bertsimas and Sym or Ben-Tal and Nemirovski’s approaches are suitable just with the assumption of independency of parameters. Is it true? Can anyone advise me about the truth of this issue?