-
assumptions made regarding variables might not be true, e.g. the distribution of supposedly normally distributed data may not be quite normal (have a back-up analysis plan!)
-
parameter space unknown (explore it a bit, use previous observations to be at least in a relevant range)
-
computational power and time (use research software engineer staff on campus to optimize code, use parallel core processing, use server services on campus, etc.)
-
simulations might be redundant with mathematical demonstrations (I don’t mind, still useful for me!)