University of Oregon, College of Education

Workshop on Effect Size and Statistical Power



This workshop focuses on effect size and statistical power.  Statistical significance testing is often misunderstood and misinterpreted.  The workshop will provide a background on fundamental concepts in Null Hypothesis Significance Testing (NHST) and then move to a consideration of how the magnitude of effects must be considered for a complete and accurate interpretation of NHST results.  Alternative measures of effect size and strength of association will be considered.  Practice will be provided in effect size calculation and interpretation.  The workshop will also define and discuss statistical power and factors that influence statistical power.  Practice will be provided in how to use software to estimate power in a variety of research designs and for different statistical significance tests. Opportunities will be provided for participants to estimate statistical power and effect size for their own research. 

Please direct any questions to Joe Stevens at stevensj@uoregon.edu.


Before the workshop, participants should:

  • Download a copy of the workshop PowerPoint slides for note taking in pdf format (one slide/page) or as multiple slides/page
  • If you want to do hands-on activities in class:
    • Download and install the free power estimation program G*Power
    • If you are interested in power in cluster randomization or HLM studies, download and install the Optimal Design Software
What you should bring to the workshop (only needed if you want to participate in hands-on activities):
  • A laptop with wireless connectivity
  • To practice applications to your own work, bring as much of the following information as is available from prior research, pilot studies, or extant research literature:
    • t-test:    Means, standard deviations, effect size, correlation between groups if dependent t
    • ANOVA:  Means, standard deviations, effect size, correlation over groups, levels, or measures in within-subjects designs
    • Multiple Regression: Number of predictors, R2
    • Multi-site or Hierarchical Studies:  intraclass correlation, R2 for predictors, number of clusters or sites, cluster or site size

           

Back to my Home Page