By Eugene Komaroff (Keiser University Graduate School)
Information
Statistical significance is still an important decision tool when working with small sample sizes. However, the concept seems to be only ritually invoked and often misinterpreted by students. For instance, they conclude the null hypothesis is true because there is no statistical significance. There will be no doubt about the truth or falseness of the null hypothesis in this presentation. Independent, identically, distributed normal, random variables with different sample sizes will be sampled from the standard, normal distribution. Consequently, approximately 5% of the statistically significant p-values will be Type 1 errors, because here it is known that the difference in population means is equal to zero. In the false null hypothesis scenario, an effect size (Cohen’s d) will be added to one of the variables. Now, the percentage of statistically significant p-values will represent power. The real-time simulation will be run with SAS OnDemand for Academics, which is no-cost, online statistical software. The author’s SAS program is available upon request so anyone can replicate the results. Because there is no set seed in the SAS program to start the random sampling stream, besides the proper interpretation of statistical significance, students will also learn the difference between reproduction and replication.