Use of the probability of type II error for determining the number of replications in the experimental designs

Abstract

This study was done on data of soil calibration curve and potassium analysis of flood rice soil. A completely randomized design was used with four levels of potassium: 0, 150, 300 and 450 kg/ha. The power test and regression coefficient were fixed to set up tables using SST and SSR to test linear hypothesis. Regression sum of square test was better than treatment sum of square in linear hypothesis, and it gave high signicant results. Fixed a desirable power test, this method provides a number of necessary replications to reach that power, this meaning substancial time and money economy.