Case and affecting factors of one-sample mean sample size estimation
Objective:To explore the rule of sample size change in one-sample means in difference and non-difference tests.Methods:The differences in sample estimation between different tests were firstly illustrated by giving examples,followed by exploring the effects of Cohen's d value,mean difference(test value-reference value,8),standard deviation(σ),and margin(Δ)and the magnitude of the ratio between them on sample size,and also exploring the effects of nonparametric correction in the one-sample mean test on the results of sample size estimation.Results:The data simulation results show a gradual decrease in sample size when Cohen's d value were taken from 0～1,and the sample size becomed smaller as it gets closer to 1,while its rate of decline also becomed smaller.When the Cohen's d value was close to 0.1,the sample size was about 1 000 cases;when the Cohen's d value was close to 0.4,the sample size was only about 50 cases;but when the Cohen's d value was greater than or equal to 1,the sample size was less than or equal to 10.Compared to the results before non-parametric correction,the sample size decreased about 32％after correction by the Double exponential method,decreased about 7％after correction by the Logistic method,and increased about 6％after correction by the Normal method,with a large difference in sample size between different nonparametric correction methods.In the non-inferiority test,the sample size increased with the absolute value of[σ/Δ]or[δ/Δ];in the superiority test,the sample size increased with the absolute value of[σ/Δ]and decreased with the absolute value of[δ/Δ];in the equivalence test,the sample size increased with the absolute value of[σ/Δ]and decreased with the absolute value of[δ/Δ].Conclusion:Clarifying the effect of parameter changes in different test statistics on sample size can help researchers clarify the pattern of sample size changes,so as to correct the values of sample size parameters according to the research purpose and research reality,and choose the correct method to calculate a reasonable and appropriate sample size.
Single sampleMeanSample size estimationEffect sizeDifference testsNon-difference testsFactors
NETL
NSTL
万方数据
