suggested threshold to reduce data noise for a factorial experiment

In this research, a factorial experiment (4*4) was studied, applied in a completely random block design, with a size of 2⁵ observations, where the design of experiments is used to study the effect of transactions on experimental units and thus obtain data representing experiment observations that the difference in the application of these transactions under different environmental and experimental conditions it causes noise that affects the observation value and thus an increase in the mean square error of the experiment, and to reduce this noise, multiple wavelet reduction was used as a filter for the observations by suggesting an improved threshold that takes into account the different transformation levels based on the logarithm of the base j and obtaining several values for the suggested threshold and applying then haar wavelet function with the cut-off hard and mid threshold and comparing the results according to several criteria.