| |
:: Course SummarySample size and power determination is a crucial step in setting up efficient R&D studies. These calculations are required to ensure that there are enough study subjects to enable the detection of anticipated effects and at the same time that resources are used adequately.
This course provides participants with the key elements and practical tools for computing sample sizes to achieve a given level of precision and power in statistical tests.:: Learning ObjectivesUpon completion of this training course, participants will be able :
To identify the factors which influence power and sample size: variability, desired precision and scope of the results, etc.
To know what information is needed to perform the calculations and how to determine it in practice
To calculate sample sizes for common study or experimental designs and under various constraints
To determine the power of statistical tests using computational tools
To use graphical tools as an aid in determining a balance between power and sample size:: Target AudienceThis statistical training session is intended for all researchers and scientists who set up experiments and studies in their work and who wish to determine the appropriate power and sample size for their tests.:: PrerequisiteThis one-day training session explains how to determine sample sizes to achieve a desired precision level for statistical tests. The course requirements are:
A working knowledge of descriptive statistics (e.g. mean, standard deviation, distributions, histograms, box-plots)
A working knowledge of hypothesis testing (e.g. t-test, z-test, significance level, Type I and II errors) and the confidence interval approach
Or, equivalently, completion of the course: Fundamental Tools in Statistics | | |
:: Topics Covered
- Introduction
- Definitions: Sample vs Population, Power
- Why Sample Size is Important
- How Low Power Can Affect Study Results
- Factors which Influence the Power of Statistical Tests
- Overview of Statistical Concepts: Hypothesis Testing, Confidence Intervals, Type I and II Error, Precision, Margin of Error, Effect Size
- Calculating Power and Sample Size
- The Necessary Inputs: Significance Level, Margin of Error/Effect Size, Estimate of Variability, Power Level
- How to Obtain an Estimate of Variability in Practice
- How To Compute a Sample Size To Achieve a Desired Precision Level for a Mean or Proportion
- How To Estimate a Sample Size To Compare Two or More Means and Proportions
- Estimating Sample Size for Other Advanced Statistical Tests
- Graphical Tools for Power and Sample Size Analysis
- Other Considerations: Post-Hoc Power Calculations, Simulation Methods, Ethical Considerations
- Conclusion
- Conventions
- Available Computational Resources: Java Applets, Statistical Software
:: Course ContentOne of the first steps in any study or experiment is to determine what the sample size should be. While the size of the sample has an important influence on the accuracy of the results, the power of the tests and the scope of the conclusions, this number is often determined ad-hoc.
This training session will provide you with the tools necessary to determine the sample size based on the study objectives in a realistic setting and by taking into account other possible constraints, such as time, resources, precision level, etc.
This training session covers the fundamental statistical concepts needed to determine the appropriate power and sample size for a variety of different statistical tests. They are illustrated with the help of examples, applications, discussions and case studies rather than the underlying statistical theory.
The course begins with an introduction of the important concepts of power and sample size, the impact of low power on the scope of statistical results and the impact of variability on power and sample size. Next, an overview of important, relevant statistical concepts is provided including a review of hypothesis testing, confidence intervals, what a margin of error represents, what effect size means and more. Two distributions are discussed: the Normal distribution for means and the binomial distribution for proportions. Statistical tools for calculating sample sizes for a variety of situations are emphasized. First and foremost, what information is needed to perform the calculations is outlined – significance level, power, margin of error or effect size and estimate of variability. How an estimate of variability can actually be determined in practice is also discussed using realistic situations. The common case of finding the appropriate sample size for a given margin of error for a mean or proportion is explained and presented in detail. This technique is then generalized for more complex cases where the study objective is to determine the appropriate sample size to compare two or more means and proportions. Examples and case studies are used extensively to explain and interpret each situation using applications from a variety of different fields. Sample size and power calculations for more advanced studies or experiments. Graphical tools for finding an appropriate balance between power and sample size are illustrated. Finally, a discussion on post-hoc power calculations, simulation techniques and ethical considerations is presented. Popular conventions and free online sample size and power tools are also provided.
| |
