Dear Colleague,

Does your system present a discontinuous phase transition? Jumps in the raw data are compatible with the existence of meta-stable states. I could not see the details, but they do not seem vertical, which implies that the system changes characteristics as localized meta-stable pieces are found out and worked out by the MC simulation.

Does "# of simulations" mean the size of the simulations? Anyway, those jumps in kurtosis, sd, and skewness happen at the same x values suggesting that your simulation is changing specific elements (which seem to be the tail elements) in your configuration distribution, leaving the central ones alone. Are your tail (distribution) elements' behavior important to your system's behavior?

Best regards.

Hi David,

Consider the following as your initial data:

X={-1, 0, 1}

Then assume your function gives this:

X={-30, -10, -1, 0, 1, 10, 30}

and accordingly, the following is the output:

X={-300, -30, -10, -1, 0, 1, 10, 30, 300}

This keeps your mean still around 0 but you may simply have jump in the other moment.