Power analysis

Question 1

type 1 error

Answer 1

when we accept a difference or relationship, but there isn’t one

Question 2

type 2 error

Answer 2

when we conclude that there is no difference or relationship, but there is one

Question 3

significance criterion

Answer 3

alpha (a) - risk of making a type 1 error (set at .05 - 5%)

Question 4

effect size

Answer 4

measure of the magnitude of a result independent of sample size

Cohen (1992)
small effect size (d = .20, r = .10) [393 p’s per group]
medium (d = .50, r = .30) [64 p’s per group]
large (d = .80, r = .50) [26 p’s per group]

Question 5

statistical power

Answer 5

probability of avoiding making a type 2 error

power is set at .80 = 20% chance

ratio of 1:4 - type 1 : type 2

depends on: 
effect size (large are easier to detect than small 
sample size (large = greater the power, smaller standard error) 
precision of measures (more reliable - better estimates - less error variance)

Question 6

prospective power analysis

Answer 6

significance criterion (a = .05)
power (.80)
effect size (estimated from previous research, pilot study, small/medium/large effect size)

= required sample size

Question 7

retrospective power analysis (SPSS)

Answer 7

effect size = sensitivity
number of participants = a priori
power of the study = post hoc

Question 8

retrospective power analysis (effect size)

Answer 8

Sensitivity

significance criterion (.05)
power (80%)
sample size known
estimated effect size

report
what kind of power analysis
number of participants
possible effect size...(d=x)
alpha (.05) and power (80%)

Question 9

retrospective power analysis (number of participants)

Answer 9

A priori

significance (.05)
power (80%)
effect size known
estimates participants needed for that effect size

report 
what kind of power analysis 
effect size was calculated for the experiment
sample needed for that effect size 
alpha (.05) and power (80%)

Question 10

retrospective power analysis (power of study)

Answer 10

post hoc

significance (.05)
effect size known
participants known
estimated power of study

report
what kind of power analysis
effect size of study 
participants used in each condition 
alpha set (.05) 
what is the power of the study?

Question 11

prospective power analysis (participants needed)

Answer 11

a priori

what kind of power analysis (prospective)
effect size (estimated, pilot study, small/medium/large)
participants needed for this
with alpha (.05) and power (80%) set