Survey research and design in psychology/Lectures/Power & effect sizes

Lecture 09: Power & effect sizes

Resource type: this resource contains a lecture or lecture notes.

This is the ninth lecture for the Survey research and design in psychology unit of study.

Outline

Explains use of, and issues involved in:

1. significance testing
2. inferential decision making
3. statistical power
4. effect sizes
5. confidence intervals
6. publication bias
7. academic integrity

Conclusions

• Decide on H₀ and H₁ (1 or 2 tailed)
• Calculate power beforehand and adjust the design to detect a minimum effect size (ES)
• Report statistical power, statistical significance, ES, confidence interval
• Compare results with meta-analyses and/or meaningful benchmarks
• Take a balanced, critical approach, striving for objectivity and academic integrity

Readings

1. Howitt and Cramer (2014a):
   1. Chapter 35: The size of effects in statistical analysis: Do my findings matter? (pp. 487-494)
   2. Chapter 36: Meta-analysis: Combining and exploring statistical findings from previous research (pp. 495-514)
   3. Chapter 38: Confidence intervals (pp. 529-539)
   4. Chapter 40: Statistical power: Getting the sample size right (pp. 562-586)
2. Wilkinson, L., & APA Task Force on Statistical Inference. (1999). Statistical methods in psychology journals: Guidelines and explanations. American Psychologist, 54, 594-604.

Handout

• Lecture slides (Google Slides)
• 2018:
  • Download 6 slides to a page:
  • Download 3 slides to a page:

See also

• Multiple linear regression II (Previous lecture)
• Summary & conclusion (Next lecture)
• Data dredging (Wikipedia)
• Funnel plot (Wikipedia)

External links

• p values and statistical significance (A New View of Statistics)
• Statisticians issue warning over misuse of P values: Policy statement aims to halt missteps in the quest for certainty (2016)
• The ASA's statement on p-values: context, process, and purpose (2016)

Power calculators

• Post-hoc statistical power calculator for multiple regression (danielsoper.com)