To obtain a probabilistic answer of high resolution and to shorten the time required for completing estimation of the device performance by processing statistically a pair of equations showing a data set corresponding to the answer of a model against the input data showing a device parameter and performing the probabilistic display of the device performance.

The input data 20 showing a device parameter are given to an experiment design (model) 22 which is produced by the computer software. The data 20 are not equal to the single input value and equal to the distribution of input value. The model 22 includes an expected range of input value, and the result of the model 22 is equal to a data set that describes a multidimensional surface showing the device performance and is expressed by a series of equations 24 which are shown as f(x). The equations 24 are arrayed on a spread sheet and accordingly a model 25 is set based on these equations 24. Then a probabilistic answer 28 including the distribution and statistic concerning each output value is obtained by means of a statistic routine (e.g. Monte Carlo analysis) 26.