Brainscape's Knowledge GenomeTM


Top Flashcards (Certified)
All Flashcards

PSYCH 315 > POWER AND EFFECT SIZE > Flashcards

POWER AND EFFECT SIZE Flashcards

1
Q

What are 3 categories included under statistical significance?

A
  1. Decision errors
  2. Power
  3. Effect size
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

What are decision errors?

A

Decisions Errors refer to the probability of making a wrong conclusion when doing hypothesis testing. It includes Type I and type II error ( alpha value and beta value)

In other words, it is when the right procedure leads to the wrong results

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

How come it is possible to make decision errors?

A

We use hypothesis testing to make decisions about populations by looking at samples. Since it’s based on probabilities, the process is designed to keep decision mistakes very low—less than 5%.

Hypothesis testing helps us decide if the patterns or differences we see in the sample are real or just due to chance.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

A lot of the time, decision errors happen because…

A

we don’t have enough power.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

What are the four possible outcomes in significance testing?

A
  1. True positive result: real effect; you correctly reject the null
  2. False negative: real effect; incorrectly retain the null (Type II error)
  3. True negative: no effect; correctly retain the null ( basically saying there is no effect, no difference)
  4. False positive: No effect; incorrectly reject the null.

** OUT OF THE FOUR, ONLY ONE CAN HAPPEN AT A TIME.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

What is a type I error?

A

It is the probability of rejecting the hull hypothesis when in fact it is true (should’ve been retained). In other words, it is concluding from the results of a study that the research hypothesis is supported when in fact the research hypothesis. is actually false.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

When can you know that you made a type I error?

A

When someone tries to replicate your study and they can’t, maybe they control confounding variables you’d dint, or different participants….

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

What is alpha?

A

the PROBABILITY of making a type I error (The same as significance level)

alpha = 0.05 = 5% probability/chance of making a type I error

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

What is a type II error?

A

The probability of retaining the null hypothesis when in fact it is false. In other words: it is concluding from the results of a study that the research hypothesis is not supported when in fact the research hypothesis is true. so hence why probability of having that sample, must be smaller than the probability of making a type 1 error.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

How are type II errors possible?

A

Even though the research hyp. is true, the power may not be large enough to detect an effect in a particular study (i.e. small sample size, so you cant have significant results:(

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

What is beta?

A

The probability of making a type II error.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

EXAM QUESTION:
Are type 1 and type 2 errors inversely related?

A

YES. If you decrease the probability of making a type I error, you increase the probability of making a type 2 error.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

What is statistical power (power)?

A

The ability of a test to detect an effect of a particular size (the probability of rejecting Ho when it is ACTUALLY false) - so corresponds to the right side (alpha) and left side is beta.

In other words, it is the probability of correctly rejecting the null hyp when it is false and the probability of correctly concluding that you 2 population means differ significantly. So basically not fucking up (making the right decision!)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

Would you rather avoid a type 1 error or type 2 error?

A

Type 1. EXAMPLE: PRODUIT FARMACEUTIQUE.

type 1: law suit
type 2: Well you lose nothing except helping la population ciblée, mais tu sais at least tu n’aggraves pas leur santé.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

How is power calculated and what is a “good” power to aim for?

A

1-Beta.

0.8 is a good level to aim for.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

What decreases the likelihood of decision errors?

A

More power.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
17
Q

2 characteristics about the relationship between type 1 and type 2 errors.

A
  1. Only one of these errors can occur in a given study
  2. They are inversely related; controlling one, increases the other. The smaller the alpha level, the higher the beta level.
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
18
Q

What distribution are we using when discussing power?

A

Distribution of means under the ALTERNATIVE HYPOTHESIS!!!!!!! not the null

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
19
Q

What is the basic concept/approach to figuring power and beta? (steps)

A

Step 1: Find the Z score on the distribution of means under the ALTERNATIVE hypothesis that corresponds to the critical mean (Sample mean (X avec barre) critical in the distribution of means under the NULL hypothesis.

STEP 2: the probability of exceeding this Z score (which is the power of the study) can then be found in the normal curve table.

20
Q

MORE POWER = …
LESS POWER = …

A

less overlap between distribution of means under the alternative hypothesis and under the distribution of means under the null hyp. So that mean there is a significant difference between the null hyp and research hyp.

more overlap between […]

21
Q

In the steps of figuring power and beta, u1 represents what?

A

The mean of the distribution of means under the alternative hypothesis (same as sample mean, a point estimate of the population mean!)

22
Q

What is the “critical mean”? What is the formula for the critical mean?

A

It represents the value of the mean corresponding to that critical Z (which in turn corresponds to the alpha level on the distribution of means under the null)

Critical mean = the mean of the comparison population (under the null) + (Z CRITICAL x Standard deviation of the distribution of means under the null hypothesis.

23
Q

If the alternative hypothesis is
H1: u1>u0 (one tail positive critical value), the proportion below this Z score is …

A

Beta, thus Power = 1- Beta

obtained test statistic is positive (proportion below obtained test statistic is beta)

24
Q

If the alternative hypothesis is
H1: u1<u0 (one-tail negative critical value - Z score), the proportion below this Z score is …

A

Power, thus Beta = 1- power.

obtained test statistic is negative (proportion below obtained test statistic is power)

25
Why do we compare a sample to a distribution of sample means and not to the population?
because the probability of getting an average is less than the probability of getting a single score
26
What is the critical mean under the null?
The critical value is a threshold on the null distribution that determines the boundary for rejecting H0 .It is based on the significance level (α).
27
when figuring out power and beta, why do we have to find the z score on the distribution of means under the alternative hypothesis that corresponds to the critical mean in the distribution of means under the null hypothesis. and then the probability of exceeding this Z score (which is the power of the study) can then be found in the normal curve table. When discussing power we are using the distribution of means
To determine the power, we calculate how much of the alternative distribution lies beyond this critical value. This involves finding the corresponding z-score in the alternative distribution. so the less the overlap, the more the power!!
28
When figuring out power and beta of a study for one tail (positive or negative critical value), what are the 4 steps to follow?
1. Gather the needed info 2. Find the characteristics (um and om) of the distribution of means under the null hyp 3. Find the critical mean in the distribution of means under the null - Z critical (alpha of... ; 1 tailed) - mean critical = u0 + (z critical)(om) 4. Find the Z score on the distribution of means under the alternative hypothesis that corresponds to the critical mean in the distribution of means under the null hypothesis. dependemment de right or left tail. beta = P(z) (right tail) Power = P(z) (left tail)
29
remember "critical" = cut-off value
:)
30
What are 5 things that influence the power of a study? Are they primary or secondary influences?
PRIMARY INFLUENCES - Effect size - Sample size SECONDARY INFLUENCES - Alpha - 1 or 2 tailed test - type of statistical tests (beyond this class!)
31
What is effect size? How can we increase effect size, and how does it influence power?
effect size is the magnitude of the difference between groups Cohen's d (Cohen effect size) is the STANDARDIZED mean difference (SMD) between two distribution units. Cohen effect size indicates how many standard deviations the populations differ by d= u1-uo/pop.standard dev. EXAM: TWO WAYS TO INCREASE IT: 1. the bigger the difference between u1 and uo, the greater effect size is (denominator goes up) 2. Pop. stand dev = spread. As it gets smaller, effect size value goes up INFUENCE ON POWER: 1. The greater the mean difference (u1-uo), the more power. WHY? - the larger the difference between the two means, the less overlap there is between the two distribution of means (they are farther apart!) 2. The smaller the population standard deviation , the more power. WHY? - The smaller the population standard dev., the smaller the standard deviation of distribution of means, and thus the less overlap between the distribution of means (they are each narrower) **** brefffff the higher the effect size, the more likely the two populations are different from each other and that's exactly what we want!!!!!!!
32
How does sample size influence the power of a study?
The larger the sample size, the more power. WHY? The variance of the distribution of means is based on the population variance divided by the sample size (o2/n ou for standard dev: o/racine de n), thus the larger the sample size, the smaller the variance, the less overlap of the distribution of means (they are each narrower) **which is why on sassure devoir un large sample size (representatif -> more power)
33
How does the alpha level influence the power of a study?
The less stringent (less strict) the significance level, the less extreme the cutoff score (more area is included in the power region), easier to reject the null (higher chances of type 1 error). So as critical z value goes up, power goes down (
34
Which tests have more power? One-tailed tests or two-tailed tests?
One-tailed tests have more power. WHY? The cutoff score in the predicted direction is less extreme/less stringent (since all of the alpha percentage is in that end instead of being divided in half)- on se donne plus de chance devour type 1 error, so less chance of having type 2 error!!! Go check screenshot iPad for summary (influences on power and practical ways of increasing power of a planned study) !!!!!!!!
35
What is an acceptable level of power for any study?
0.70 (70%)
36
Standard tables for directly determining the power of a study require sample size and effect size. (this is nice, but you know in reality we have no idea of the population parameters so keep that in mind)
So you know you can either do all the previous steps we saw for finding beta and power orrrrrrr, you find the effect size (Cohen d) and with the sample size, and the right table (either a one tailed test or two tailed tests) you find the power!! (take the sample size closer to your real sample size, if you're actual sample size in not included in the table!!!) - btw power tables assume alpha = 0.05
37
What type of error refers to the probability of saying pop means do not differ significantly when in fact they do?
beta
38
When are power tables also useful?
When planning studies, helps you identify how many participants you will need in your study to achieve an acceptable level of power (i.e. 80)
39
4 examples of conclusions. Outcome... statistically significant (small and large sample) NOT statistically significant (small and large sample)
Outcome statistically significant small: important results large sample: might or might not have practical significance small: inconclusive large sample: research hyp is probably false (unlikely to be true based on the evidence we have)
40
During exams, do we always got heck the power table for a power question?
No, always do the full steps EXCEPT when he specifies to use the table, and that is usually when we want to estimate best sample size.
41
effect size is measured in...
standard deviations!
42
Cohen's d is the same as a...
z score of a standard normal distribution. Using cohen's d can be converted into a scale of percentile between two compared groups known as "Cohen's U(petit 3)" So based on the value of Cohen'd , we can get the proportion of control group which would be below the mean of the treatment group, and thus which would tell us about the size of the effect (small, medium or large)
43
between 0 and 0.4 (effect size), size of effect is..
small effect
44
between 0.4 and 0.8 (effect size), size of effect is..
medium effect
45
From 0.8 (effect size), size of effect is...
Large effect. think about it effect size= 3, means differ by 3 standard dev. so overlap ridiculously small!!

PSYCH 315 (9 decks)

Brainscape helps you reach your goals faster, through stronger study habits.
© 2025 Bold Learning Solutions. Terms and Conditions