Brainscape's Knowledge GenomeTM

Browse over 1 million classes created by top students, professors, publishers, and experts.


See full index

psy3062 > Interaction and moderation in Linear Regression > Flashcards

Interaction and moderation in Linear Regression Flashcards

1
Q

What is a main effect?

A

The effect of one factor (IV) on its own

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

What do ANOVA’s do in terms of effect?

A

They examine the effect of each factor on it’s own (IV), AND also examines the combined effects of the factors

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

What is the difference between a main effect and an interaction?

A

An interaction examines two or more factors at the same time - their combined effect, which may not be predictable based on the effect of either factor on their own

A main effect is the effect of one factor on its own

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

Why do Factorial ANOVA’s produce multiple F ratios?

A

There is one for EACH main effect an interaction term.

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

Do linear regression with moderators produce main effect terms?

A

Produce main effect terms AND an interaction term, all with own coefficients

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

Does it matter if an interaction is continuous or categorical?

A

No

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

What is the difference between what an ANOVA asks versus what a Linear Regression asks?

A

In an ANOVA a predictable variable is a factor (groups of subjects) , and it is asking what is the difference in DV BETWEEN groups whereas a linear regression with an interaction is asking how MUCH (continous) does the DV change when the IV changes THIS much.

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

Moderation is used more in _____ and Interaction is used more in ________.

A

Moderation is used to describe a second IV in regression that moderates another IV

Interaction is used for factors in ANOVAs

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

In both regression and ANOVA, an interaction between two variables means….?

A

That the association between one IV and DV DEPENDS ON (is moderated by) the other IV

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

If you have an interaction between two variables, and it appears that the second IV strengthens the association between one IV and DV, what type of moderation is this?

A

Enhancing moderation

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

If you have an interaction between two variables, and it appears that the second IV reduces the association between one IV and DV, what type of moderation is this?

A

Buffering moderation

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

If you have an interaction between two variables, and it appears that the second IV both increases and reduces the direction of the association between one IV and DV, what type of moderation is this?

A

Antagonistic moderation

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

Which of the following applies to moderation?

A. Moderation refers to condition where the effect or relationship of a predictor with the outcome depends upon another variable
B. Moderation implies that the relationship between two variables is not the same for everyone
C. Moderation is in linear regression only
D. Moderation is used for factors in ANOVA and interaction in regression

A

A and B

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

Which of the following is not a type of moderation?

A. Enhancing
B. Covariate
C. Buffering
D. Antagonistic

A

B

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

What do we call the variable that modifies the relationship being studied?

A

THe moderator

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

What type of analysis has this research question most likely conducted?

Coping buffers the effects of stress and mental health problems

A

Moderation

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

What type of analysis has this research question most likely conducted?

People who have experienced some stress in life are more resilient to future stress

A

Modertional

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

What type of analysis has this research question most likely conducted?

Having at least one warm parental figure (e.g., mother, father, aunt/uncle, grandparent) is protective against the deleterious effects of childhood trauma on outcomes in adulthood

A

Moderation

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

What type of analysis has this research question most likely conducted?

Cognitive behavioral therapy improves sleep when the source of poor sleep is psychosocial (e.g., insomnia) but not when it is due to disruption (e.g., noise, nighttime pain)

A

Moderation

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
20
Q

How is moderation tested, statistically?

A

Through interactions

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
21
Q

What are interactions?

A

Combinations between two or more variables

They can be between two categorical variables/continous variables / or a categorical and continous variable

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
22
Q

In regression, interactions are added as additional variables/terms. How is this done?

A

By multiplying the two IVs after mean centering, to get a NEW interaction term

23
Q

Do interaction terms have their own new coefficient?

A

Yes

24
Q

Do interaction terms have their own new coefficient?

A

Yes

25
Q

In the ineraction equation, what else does the coefficient refer to?

A

The slope

26
Q

True or false: like in correlation, the moderator and predictor are reversible

A

True

27
Q

If you have multiple hypotheses, and you want to see if a one variable effect on the DV depends on the effect of another variable, what would you use?

A

Heirarchical LR.

28
Q

IF you have four variables interacting, what type of interaction is this called?

A

Four-way interaction

29
Q

Are interactions additional variables added to a model?

A

Yes.

30
Q

What does the b1 and b2 represent in an equation for multiple regression?

A

The main effects

31
Q

In an interaction equation with continous variables, what is b naught?

A

The estimated OUTCOME when Vairable X1 and Variable X2 = 0

32
Q

In an interaction equation, what is b1?

A

B1 is the estimated difference in Outcome for a 1 unit change in Variable X1 when X2 is zero (the simple effects of x1)

33
Q

In an interaction equation, what is b2?

A

B2 is the estimated difference in Outcome for a 1 unit change in Variable X2 when X1 is zero (the simple effects of x2)

34
Q

What is b3?

A

The estimated difference in b1 slope when variable x2 changes 1 unit OR difference in b2 slope when variable x1 changes 1 unit

35
Q

If you have a interaction with a continous and categorical variable, what do we need to do to the categorical variable?

A

Dummy code it. So three categories
Have one reference group
Then two dummy variables

36
Q

If we have these three categories : Highschool, some uni, uni grad, and we want to create an interaction term with them, how would we do this?

A 
First dummy code. 
Put into three categories:
SomeUni
D_Grad
HS

then, create two interaction variables (not just one)

37
Q

Say you have a continous predictor variable and a cateogrical predictor variable.

How many main effects would you have with two predictor variables AND two interaction terms - because you had two levels of a categorical predicotr?

A

Well you would have three main effects from original equation due to two levels of categorical predictor (ignore reference group) and other continous variable. So that’s three

But then you’ve created two extra interaction terms using categorical predictors on continous variable.

So 5 in total.

So intiially two IVS became three main effects with cateorgiacl predictor

then two itneraction terms

38
Q

What do continous variables that interact with itself allow s to test?

A

A quadratic relationship between a variable and an outcome. So non linear associations

39
Q

If you want to test a quadratic relationship, what type of interaction or variable within that interaction would you want to assess?

A

A continous variable that interacts with itself

40
Q

Why would we have a continous variable interact with itself ?

A

Because relationships are not truly linear all the time and we might see simple slopes are different at different values of the variable, depending on where you are on the x axis. Direction does not really matter but the magnitude.

41
Q

If you can see the slopes are not the same at each point on an x axis, is the association linear?

A

no

42
Q

If you decide to mean centre your varibale, and have 20 the centre, with original point X=5, what would the new value become?

A

5-20 = -15

43
Q

Is mean centering the same as standardising?

A

No becuase, standardising makes each unit a change in standard deviation.

44
Q

Do we need to center dummy coded variables?

A

No because zero is already meaningful (the reference group)

45
Q

What should we do post hoc when reporting simple slopes with two continous IVs?

A

To sum up whole reporting simple slopes thing, basically if you got two continuous IV and X is your predictor variable and M is your moderator, then what you should do post hoc, maybe might have aprior, but usually doing post hoc, is follow it up with simple slope of X for a low value of M, and again usually define Low as the mean of the moderator minus 1 SD, and then also show simple slope of X at high value of moderator. If plotting things prob also plot simple slope at just mean of moderator so then showing three regression lines – one for each slope at high, low and average levels of the moderator. If moderator categorical, even easier as just calculate simple slope at each level of the categorical variable so if you have biological sex as you categorical moderator then calculate a simple slope for male participants and simple slope for female participants.

46
Q

If you have two categorical variables and a outcome that is continous, can you do a moderation with multiple regression?

A

You should be looking at a Factorial Anova.

Your outcome variable (Y) should be continuous, and need to have one IV (the X be continuous) its ok if M is categorical, but if both are categorical (IVs) and no other continuous variables in multiple regression, you should be thinking of it as factorial ANOVA.

47
Q

Are assumptions the same for moderation linear regression and correlation?

A

Yes. Also outcome values or data points should be independent of onservations
DV is continous
One IV is continous
No multicollientarity
No significant outliers
Assumptions of linearity - association between X and Y should be linear. If M is continous, association between M and Y should be linear. Association between X x M and Y should be linear. And if M is categorical, can visually look for non linear associations

48
Q

Can moderators give causal inferences?

A

No but can suggest possible mechanisms/mediators.

49
Q

What is the difference between moderators and mediators?

A

Moderators are the levels at which the association between two other variables depends on.

A mediator is something where the effect of X on Y goes THROUGH the mediator – that would be X CAUSING the mediator, and then the MEDIATOR in term causes why.

Moderation good for pointing out individual differences.

50
Q

When do we need to be extra careful with multicollinearity between X and M ?

A

When M is a categorical because may not be able to tell is association between X and Y is non linear..

51
Q

With categorical moderating variable, the category with smaller variance is expected to have a weaker or stronger slope?

A

Weaker

52
Q

IMPORTANT POINT: When assessing assumptions, if your moderator is significant, what should we always ask of M - is it changing the relationship between X and Y or is it changing the variance in ____?

A

X

Example: If association between say stress and satisfaction depends on age, maybe being older means stress and satisfaction are more strongly related, or maybe the effects are actually the same but there is just more variance in stress when you are younger. Visually inspect homogeneity of variance separately by groups, same way… just read last line.

53
Q

How do we assess homogeneity of variance in moderated linear regression?

A

By visually inspecting homogeneity of variance (in residuals VS PREDICTED plot) separately by groups the same way we have inspected for single predictor variables

psy3062 (9 decks)

Brainscape helps you reach your goals faster, through stronger study habits.
© 2024 Bold Learning Solutions. Terms and Conditions