Probability of finding an effect given that one truly affects
1 of 37
What is P?
The probability of observing data given that H0 true power = probability of finding P
2 of 37
What is H0?
null hypothesis
3 of 37
What is N1?
Alternative hypothesis
4 of 37
If H0 in the experiment is accepted and it is in real life what is this called?
A true negative
5 of 37
What if H1 is real life but H0 is what the experiment is showing?
Type II error probability = B
6 of 37
What happens if the experiment is accepting H1 but in real life the H0 is accepted?
Type I error false alarm probability = A
7 of 37
What is H1 is accepted in the experiment and in the real world?
True positive probability = 1-B
8 of 37
What is a type I error?
When the null hypothesis is true and you reject it
9 of 37
What is type II error?
When the null hypothesis is accepted but it is actually false
10 of 37
What does power show?
Probability of finding an effect given that there actually is one
11 of 37
What will power be governed by?
Size of the effects, number of subjects and choice of alpha, sources of variability, experimental design, choice of test, tais of a test
12 of 37
How is a size of effect affected?
Experiment: metal Vs wooden ball dropped from top of tower, if metal ball genuinely falls faster than wooden one, there should be a better chance of measuring the difference in their speeds if it is lare
13 of 37
How does a number of subjects affect significance testing?
If there is a very small effect then more participants may be needed to be sure that the observed effect is real
14 of 37
For example?
We need to take more samples were more confident that we know the real means, with greater confidence in our knowledge of the means comes greater confidence that we can spot a difference
15 of 37
What does the choice of alpha depend on?
different outcomes of false positive and false negatives
16 of 37
For example?
If the outcome falsely rejected the null hypothesis is bad, use low alpha
17 of 37
What happens if the outcome of falsely accepting the null hypothesis?
It is bad then we have to use high alpha
18 of 37
What is standard alpha?
0.05
19 of 37
What is the source of variability?
WE need to minimise noise variables, experimenter must consider whateer may influence subjects response which is irrelevant to the study
20 of 37
For example?
Present faces of different emotions and measure resonse time to detect each one
21 of 37
What is the problem?
RT involves two components time to perceive and time to make a response we are not interested here in the latter
22 of 37
What is the solution?
Present the faces for a vriable length of time and see how long is enough for the percetion to be accurate --> we removed additional noise
23 of 37
What is the experimental design?
A particular source of variability is individual differences, where possible/practical try to use a within subject design
24 of 37
What is a two tailed test?
We hypothesised that subjects scores in condition A will be different to those in condition B
25 of 37
What is a one tailed test/
We hypothesis that subjects score in condition A will be greater than those in condition B
26 of 37
What are tails of test?
suppose that H0 was actually true and we ran our experiment, 1 million times
27 of 37
What would we presume?
The liklihood our data would come about assuming that the null hypothesis is true
28 of 37
What is the one tailed test?
P value is the probability of seeing this big difference in a given direction
29 of 37
What is alpha?
The 5% of the total area is split between the two tails
30 of 37
What would happen if P
2 tails: harder to find an effect
31 of 37
What if P
A very clear prediction in one direction
32 of 37
What are one tailed tests?
More powerful than 2 tailed tests
33 of 37
What are one tailed tests based on?
Some prior knowledge if you predict an effect in one direction and then one in the other, with a one taield test, you could conclude there s no difference
34 of 37
What is the required information?
Size of effect, number of subjects and choice of alpha
35 of 37
what is a family wise error?
Inflated type I error
36 of 37
What is the correction for multiple comparisons?
The bonferroni correction jus divides the alha value by the number of stats comparisons made
37 of 37
Other cards in this set
Card 2
Front
What is P?
Back
The probability of observing data given that H0 true power = probability of finding P
Card 3
Front
What is H0?
Back
Card 4
Front
What is N1?
Back
Card 5
Front
If H0 in the experiment is accepted and it is in real life what is this called?
Comments
No comments have yet been made