Statistical power may be understood as being an indicator of the ability of a test of significance to spot a practical difference, such as the one between the averages of two products that are being compared. In this context, a low power typically means that the sample sizes in the study are too small. Conclusion of what is “non-significant” is justifiably questionable when it is arrived at without an analysis of statistical power. For this reason, unless power is high, a study may be doomed to failure even before it is begun.
Interpreting the power analysis
So, a complete understanding has to be made of how to interpret and use the power-analysis outputted by text-book calculations or software programs modules, such as StatgraphicsCenturionXV.
A webinar that is being organized by Compliance4All, a very popular provider of professional trainings for all the areas of regulatory compliance, will throw light on these aspects. John Zorich, Statistical Consultant & Trainer at Ohlone College & SV Polytechnic, will be the speaker at this webinar. To enroll for this learning session and enhance the understanding of statistical power, just visit http://www.compliance4all.com/control/w_product/~product_id=501048
A test of statistical significance, when it is conducted with the hope that the result will be non-significant, may be unacceptable to a regulatory agency unless the test had an acceptable level of “power”. The FDA typically requires a minimum of 80% power, and often requires something like 90% power.
However, the calculation of this power is not too easy. It is often so complicated that it typically needs a software program to carry it out. Even with all this, the software program’s output can be misunderstood if the user does not have a proper understanding of the basic concept of statistical power.
Demonstrating with examples
At this learning session, John will explain the basics, by using a t-test as an example. He will demonstrate one of the several possible formulas, as well as two different software programs and their “Power Curves”. Professionals who deal with statistics, such as QA/QC Supervisor, Process Engineer, Manufacturing Engineer, QC/QC Technician, Manufacturing Technician, and R&D Engineer will find this session highly useful.
John will cover the following areas at this webinar:
- Vocabulary and Concepts
- t-Tests and p-values
- Statistical power
- for t-Tests
- critical difference to detect
- example calculations
- Power Curves