W5 - 1-Way Independent ANOVA Flashcards

1
Q

Define experimental power

A

Ability to detect an effect of a certain size

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

What do underpowered studies leave you unable to do?

A

Detect small effects

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

What can you do to ensure the study has enough power?

A

Test plenty of people

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

Type 1 error

A

Incorrect rejection of the true null hypothesis

When 1 concludes a relationship or effect exists when it actually doesn’t.

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

What is the level of type 1 error we are willing to risk called?

A

Alpha

Type 1 errors will occur 5% of the time (1 in 20).

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

Type 2 error

A

Failure to reject false null hypothesis

When 1 concludes a relationship or effect doesn’t exist when actually it does.

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

What is the level of type 2 error we are wiling to risk called?

A

Beta

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

What are statistical tests based on the assumption of?

A

That variance is due to true and random effects

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

List reasons as to why you may have a p value larger than 0.05

A

No actual effect

Lack of power - Too small a sample to detect an effect of a certain size

Inadequate variability w/in IV

Measurement error

Nuisances variance

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

List some sources of measurement error

A

Inadequate measurement instruments

Response error from participants

Contextual factors

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

Equation for family wise error

A

1-(1-alpha)^no. tests conducted

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

What happens to familywise error with the increasing number of tests we run

A

More inflation of familywise error rate

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

How can the increased family wise error/error rate be corrected?

A

By the bonferroni, Tukeys HSD, Holm & Scheffe tests

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

What do the bonferroni, Tukeys HSD, Holm & Scheffe tests do?

A

Reduce your ability to detect a true effect

Lower experimental power

Increase chance of making a type 2 error

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

What is a solution to the bonferroni, Tukeys HSD, Holm & Scheffe tests lowering experimental power + increasing chance of making a type 2 error?

A

1-way independent ANOVA

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

What does the 1-way independent ANOVA allow you to do?

A

Compare many groups w/ a single test

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

Positives to a 1-way independent ANOVA

A

No inflated familywise error rate

No reduction in experimental power

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
18
Q

Assumptions to the 1-way independent ANOVA

A

Samples are independent

Data is interval or ratio

Normally distributed

Homogeneity of variance

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

What is the variance in ANOVA tables?

A

Mean Square (MS)

20
Q

What does ANOVA ask the question of?

A

How successful the experimental manipulation has been in comparison to the natural variability that exists.

Compares the type of variance you want (effect variance) with the variance you do not want (error variance).

21
Q

What is the F ratio

A

Ratio between wanted (effect) variance and unwanted (error) variance.

22
Q

What does the 1-way independent ANOVA do specifically?

A

Examines the ratio of variance between groups + variance within groups

MS between / MS within

23
Q

When do significant effects occur in the 1-way independent ANOVA

A

When you have 4 times as much variance between groups than within.

24
Q

What does an F ratio close to 1 mean?

A

Treatment creates little variability in scores

25
Q

When can a significant difference be shown in regards to the F ratio

A

When theres a large F statistic but depends on the df

26
Q

Can the F ratio ever be negative?

A

NO

27
Q

Is the F ratio ever less than 1

A

Hardly ever

28
Q

What size are most F ratios

A

Medium

29
Q

What does a small F ratio usually indicate

A

No effect

30
Q

What does a large F ratio usually indicate?

A

A sig effect but depends on df

31
Q

How can you find out if the F ratio is sig?

A

Calculate df between groups (effect/treatment)

Calculate df within groups (Error)

Find values on critical value table.

If F value is larger than the critical value for our df then there IS a sig difference at the p<0.05 level.

32
Q

Why is error variance (a.k.a measurement error) always present?

A

Due to variables not being able to be measured perfectly + there being ind differences

33
Q

Error variance (MSE) in independent ANOVA

A

Always occurs within each group (MSwithin)

34
Q

Calculating the F ratio

How to calculate the SS within

A

Mean of each group

Subtract mean from a dataset from each value in that dataset.

Square the differences

Total the squared differences to get the SS within.

35
Q

Calculating the F ratio

How to calculate the SS between

A

Sum of means

Square of means

Total the squared means

Calculate SS for means of the groups (SSm)

Multiply SSm by no. of ppl in each group to give the between group effect (variance)

36
Q

What must happen if the F value is greater than the critical value of 3.68?

A

Reject null hypothesis as there IS a sig difference at the p<0.05 level

Meaning 1 of our mean values is sig different from 1 of the other sample means. but don’t know which as theres more than 2 groups so we need to carry out a POST HOC TEST

37
Q

What are post hoc tests used to determine?

A

Which pairs of means are different from 1 another

38
Q

List the post hoc tests

A

Tukeys HSD test
Bonferroni Correction test - Usually this one
Scheffe test
LSD test

39
Q

How is effect size reported?

A

Partial eta squared

40
Q

Define effect size

A

% of variability in the DV explained by the IV.

41
Q

Which post hoc tests provides the most conservative correction?

A

Bonferroni

42
Q

What does the Bonferroni post hoc test do?

A

Counteracts inflated familywise error by creating a new alpha.

Does this by dividing original alpha by no. of tests conducted.

43
Q

When must you avoid using the Bonferroni post hoc test

A

When theres lots of conditions

44
Q

What does the Tukey HSD post hoc test do?

A

Calculates a new critical value

45
Q

When should the Games-Howell test be used?

A

When you have unequal variances between groups due to levenes test p value being <0.05.

46
Q

What does the LSD post hoc test stand for?

A

Least sig difference

Should be avoided