Power (concepts and calculating)

Question 1

What is power

Answer 1

Probability of correctly rejecting a false H0 (carried out before study

Question 2

Factors affecting power

Answer 2

1. size of alpha (decreasing alpha = decreased power, increasing alpha = increased power)
2. directionality
3. size of effect (bigger size of effect = easier to detect (power)
4. size of population standard variation (affects SEM) - (as sigma increases, SEM increases, power decreases)
5. size of number of participants (affects SEM) - as the number increases, SEM decreases, power increases)

Question 3

What is needed to calculate power

Answer 3

• ⍺
• 1-tailed of 2-tailed
• 𝛾
• σ
• n

*Cannot calculate ‘the’ powder for an experiment - it requires knowing the true effect size 𝛾 which depends on the true population mean (μ1)

Question 4

Cohen’s rule of thumb

Answer 4

• Used when researcher doesn’t know enough about the DV to identify the effect size

size of effect
.2 = small effect
.5 = medium effect
.8 = large effect

Question 5

Effect size

Answer 5

An effect size is a standardized measure telling you how big to expect the mean differences between conditions to be

Question 6

What is fixed in power

Answer 6

gamma, alpha and the tails are typically fixed the only thing variable and manipulable is the number of participants and the power

Question 7

How does confidence interval show precision

Answer 7

• Precisions is reflected in the width of the confidence interval
  ○ Higher precision = narrower interval
  Increasing n = increased precision

Question 8

What are the methodological issues when calculating power

Answer 8

○ Draws attention to β (type 2 error) and 1-β (power or sensitivity)
○ Focus on effect size as well as significance
○ Highlights the distinction between statistical vs practical significance

§ By specifying a minimum effect size of interest = saying that any smaller effect than that value has so practice important (even if H0 is false)
§ All null hypotheses can be said to be false since true value of μ is unlikely to be exactly equal to μ0
Hence focus should be on whether H0 is false to an important degree (minimum effect size of practical importance)

Question 9

As the P value decreases….

Answer 9

P value decreases as n increased because it reduces the standard error or variance and increases t stat

T increase p decrease