EN
In the statistical process control the assumption of normal distribution is widely used. But in practical this assumption is often unfulfilled. In this situation appropriate transformation or nonclassical method should be used. The most commonly used methods of transformation are the curves Johnson. The most important problem with the use of Johnson curves is estimation of parameter and form of curves. This paper presents improvement of Slifker and Shapiro procedure (1980) by construction of optimization model based on chi-square statistic.