Data from a production process usually is correlated and doesnot fit normal distribution. In order to detect the existence of strutural changes in production process, especially attribute data, which is focused on changes that are influenced by the data covariance structure, it is necessary to model the covariance function which identical to the spectral distribution first. Accordingly, data is transformed into its spectral distribution by using Walsh-Fourier Transformation so that data will not be correlated and can be analize statistically. Transformed data will be tested with F-test to see if this method can detect the changes. This simulation will use time series data which is generated with INAR (Integer Valued Auto-Regressive) model.