I have hundreds of parameter sets such as
set1: [p_1_1, p_1_2, p_1_3, ..., p_1_N];
...;
setM: [p_M_1, p_M_2, p_M_3, ..., p_M_N]
which I use to run simulations of an ODE based biochemical model (mechanistic drug action model). Few of them, about 10%, cause the simulations crash, most likely due to stiffness in the model.
So it could be parameter set 3, 7 & 21 and for other drug regimen yet another collection of parameter sets, e.g. 3, 78, 299. Y=There is no obious pattern in the crashed sets but it looks like the model is sensitive to certain parameter value combinations.
What would be the best method to find most important parameters causing the model to crash? I was thinking about clustering, whixch I am going to perform next but is there anything else I should try?