The main issues in the design and optimization of a chemical process under uncertainty are the feasibility test, flexibility index and the two-step optimization problem. The formulations of these problems are based on the implicit supposition that at the operation stage it is possible to correct all uncertain parameters in the chemical process models. However, in practice, one can improve the accuracy of only some of the uncertain parameters. As a result, we consider two groups of uncertain parameters. We postulate that one can significantly improve the accuracy of the first group of uncertain parameters using process data at the operation stage. This necessitates the formulation of a new feasibility test and the associated two-stage optimization problem. We consider the general case when both groups of uncertain parameters may or may not be statistically dependent. We propose methods of solving the problems based on the supposition that the parametric uncertainty region is small. Under this condition, the process models are well represented by linearized models whose accuracy is asymptotically equal to that of the original process models. This leads to substantial computational savings. Computational tests show that when it is not possible to improve the accuracy of some of the uncertain parameters, using standard flexibility analysis can lead to processes that are not flexible.