How to Determine Permissible Error δ Value Properly when Computing Sample Sizes in Hypothesis Tests

Objective There are three ways to determine permissible error δ value when researchers cannot get the difference δ between population parameters in computing sample sizes of hypothesis tests. Methods Take one-sample t-test for example and discuss how to determine δ value properly to compute the final sample size on two sides when there exists professional difference δ or not. Results Compare δ from pilot experiment and professional difference δ when professional difference δ does exist and then compute sample size; when there is no professional difference δ, we set the interval（0.25S,0.50S） to compute sample size; the results can also be applied in situations which involve permissible error like two independent sample t-test and χ² test. Conclusion It is necessary to select δ value reasonably in different situations when computing sample sizes; a flow chart for reference in practice is provided.