Week 5

• Beyond significance


  Psychologists ask a research question focused on whether a given effect is significant or not

  This approach overlooks:

  1. uncertainty/certainty of the answer
  2. how large (important) is the significant effect

  a p value tells us whether two datasets differ, but not by how much
• Confidence intervals (CIs)


  allows us to measure uncertainty about estimates

  given probability of 95% (of capturing the true population parameter)

  describes the range of values/intervals within which the true population parameter will be contained (expected)
• Effect size


  gives us a standard way to quantify the effect

  quantifying how strong the effect observed is

  • defining the size of an effect quantified from actual data
  • (+)standardized = comparable across studies
  • (+) allows others to objectively evaluate the size of observed effect
  • (-) not as reliant as sample size
• CIs: EXAMPLE

  
• 95% CI

  
• Calculating CIs

  
• Normal distributions

  
• Calculating CIs for %

  
• T distribution


  slightly broader

  width widest for smaller sample numbers

  adjusting CIs to make them slightly wider/broader/less uncertainty

  
• Calculating CI in small samples

  
• One vs two tailed hypotheses (directional effect)

  
• Sample size vs SEM and CI

  As sample size decreases, SEM decreased

  As sample size increases, CI gets narrower

  
• Measuring effect size

  further away the two distrubutions, larger Cohens d, larger effect size