*please keep in mind that this is just a summary, to pass this course I highly advise
going to the tutorials and doing ALL the additional assignments etc. that the teacher
provides for you, it really helps!!
APPLIED METHODS AND STATISTICS YEAR 3
Lecture 1: introduction
Path analysis: can you explain the correlations between the variables because of a casaul
relationship?
Factor analysis: can the correlations between a group of variables be explained by 1
underlying construct?
Structural equation modelling: can the correlations between a group of variables be
explained by underlying constructs and causal effects between those?
—------------------------------------------------------------------------------------------------------------------------
----
Characteristics of path analysis
Variables
Variables are characteristics of research units (your topic) that you are interested in
There must be variation in the characteristics across the research units
If there is not variation in the characteristics across the units, it is not a variable but a
constant
Mistakes
o Confusion of the values of the variable with the variable itself
E.g., rich and poor are two values of the same variable income
o Confusion of a process or theory with a variable
E.g., attribution theory as a variable in the path model
Relations
, A statement in which
o At least 2 variables occur
o Higher or lower values on one variable are associated with lower or higher
values on the other variable
Types of statements
o Covariation statement
o Causal statement: higher values on one variable CAUSE change in the other
variable
If you change the independent variable, then the dependent variable
will also change
Spurious relations
Causal statement 1: chocolate causes happiness
Causal statement 2: chocolate causes a long life
o Therefore: happy people live longer (covariation)
Statement 1: variable x causes y1
Statement 2: variable x causes y2
o Covariation: y1 is related to y2 (through x)
Example: on days when there are more ice creams sold, there are more shark attacks
They are both caused by hot weather
o So, they are related through a spurious relation: hot weather causes them
both
Direct and indirect effects
Statement 1: feeling blue leads to neglecting self-care
Statement 2: neglecting self-care leads to bad health
So: a valence of emotion has a direct effect on self-care, and self-care has a direct
effect on health, so valence of emotion has an indirect effect on health (via mediator
self-care)
Unknown effects
Sometimes we do not make a statement about the direction of an effect
We then simply include the correlation in the pathway with a double arrow
One direction → spurious relation by x
Other direction → indirect effect via x
Reciprocal effects
Health causes happiness
Happiness causes health
, o Two direct effects, so 2 arrows
o Reciprocal effect is often not explicitly mentioned or labelled as such
Conditional effects
Sometimes a variable does not only affect another variable, but an effect
Then, this variable is a moderator of that effect
o A mediating variable (or mediator) explains the process through which
two variables are related, while a moderating variable (or moderator)
affects the strength and direction of that relationship
Q: Stress induced increases in dopamine secretion are thus thought to explain
enhanced reward learning. What type of relation is described here?
1. Spurious relation
2. Indirect effect
3. Covariation
4. Conditional effect (moderation)
Lecture 2: path model
Relations between variables
Covariation (correlation)
o Values for both often co-occur
E.g., happy people live longer
o Often described with related to, associated with, often also, co-occur
Causal effects
o A change in one variable causes a change in another variable
E.g., chocolate makes you happy
o Often described with induce, produce, cause, affect, influence, has an effect
on, makes more likely, leads to, because
—------------------------------------------------------------------------------------------------------------------------
----
Theory → path diagram
Make a list of variables
Establish the causal order
Formulate causal hypothesis
, —------------------------------------------------------------------------------------------------------------------------
----
Testing causal hypothesis
In practice we do not observe whether a causal hypothesis is true, only whether two
variables go together. This does not necessarily imply causation because a spurious
relation can also be possible
o So, you can never prove a causal hypothesis
o So, we include both the direct effect that we are interested and the spurious
relation that may be an alternative explanation for the covariation between y1
and y2, to make sure that the path analysis can tell us which of the two
explains the covariance
Golden rule: all variables that may cause a spurious relation between two variables
with an assumed causal effect between them, must be included in the model
o Necessary extensions of model: adding common causes
Example: children are also influenced by parental pressure
So, you need to include the parental pressure variable because it can have multiple
outcomes
Causal hypotheses cannot be proven with correlational, but they can be falsified
o Causal hypothesis is rejected if: size of spurious relation is the full correlation
going to the tutorials and doing ALL the additional assignments etc. that the teacher
provides for you, it really helps!!
APPLIED METHODS AND STATISTICS YEAR 3
Lecture 1: introduction
Path analysis: can you explain the correlations between the variables because of a casaul
relationship?
Factor analysis: can the correlations between a group of variables be explained by 1
underlying construct?
Structural equation modelling: can the correlations between a group of variables be
explained by underlying constructs and causal effects between those?
—------------------------------------------------------------------------------------------------------------------------
----
Characteristics of path analysis
Variables
Variables are characteristics of research units (your topic) that you are interested in
There must be variation in the characteristics across the research units
If there is not variation in the characteristics across the units, it is not a variable but a
constant
Mistakes
o Confusion of the values of the variable with the variable itself
E.g., rich and poor are two values of the same variable income
o Confusion of a process or theory with a variable
E.g., attribution theory as a variable in the path model
Relations
, A statement in which
o At least 2 variables occur
o Higher or lower values on one variable are associated with lower or higher
values on the other variable
Types of statements
o Covariation statement
o Causal statement: higher values on one variable CAUSE change in the other
variable
If you change the independent variable, then the dependent variable
will also change
Spurious relations
Causal statement 1: chocolate causes happiness
Causal statement 2: chocolate causes a long life
o Therefore: happy people live longer (covariation)
Statement 1: variable x causes y1
Statement 2: variable x causes y2
o Covariation: y1 is related to y2 (through x)
Example: on days when there are more ice creams sold, there are more shark attacks
They are both caused by hot weather
o So, they are related through a spurious relation: hot weather causes them
both
Direct and indirect effects
Statement 1: feeling blue leads to neglecting self-care
Statement 2: neglecting self-care leads to bad health
So: a valence of emotion has a direct effect on self-care, and self-care has a direct
effect on health, so valence of emotion has an indirect effect on health (via mediator
self-care)
Unknown effects
Sometimes we do not make a statement about the direction of an effect
We then simply include the correlation in the pathway with a double arrow
One direction → spurious relation by x
Other direction → indirect effect via x
Reciprocal effects
Health causes happiness
Happiness causes health
, o Two direct effects, so 2 arrows
o Reciprocal effect is often not explicitly mentioned or labelled as such
Conditional effects
Sometimes a variable does not only affect another variable, but an effect
Then, this variable is a moderator of that effect
o A mediating variable (or mediator) explains the process through which
two variables are related, while a moderating variable (or moderator)
affects the strength and direction of that relationship
Q: Stress induced increases in dopamine secretion are thus thought to explain
enhanced reward learning. What type of relation is described here?
1. Spurious relation
2. Indirect effect
3. Covariation
4. Conditional effect (moderation)
Lecture 2: path model
Relations between variables
Covariation (correlation)
o Values for both often co-occur
E.g., happy people live longer
o Often described with related to, associated with, often also, co-occur
Causal effects
o A change in one variable causes a change in another variable
E.g., chocolate makes you happy
o Often described with induce, produce, cause, affect, influence, has an effect
on, makes more likely, leads to, because
—------------------------------------------------------------------------------------------------------------------------
----
Theory → path diagram
Make a list of variables
Establish the causal order
Formulate causal hypothesis
, —------------------------------------------------------------------------------------------------------------------------
----
Testing causal hypothesis
In practice we do not observe whether a causal hypothesis is true, only whether two
variables go together. This does not necessarily imply causation because a spurious
relation can also be possible
o So, you can never prove a causal hypothesis
o So, we include both the direct effect that we are interested and the spurious
relation that may be an alternative explanation for the covariation between y1
and y2, to make sure that the path analysis can tell us which of the two
explains the covariance
Golden rule: all variables that may cause a spurious relation between two variables
with an assumed causal effect between them, must be included in the model
o Necessary extensions of model: adding common causes
Example: children are also influenced by parental pressure
So, you need to include the parental pressure variable because it can have multiple
outcomes
Causal hypotheses cannot be proven with correlational, but they can be falsified
o Causal hypothesis is rejected if: size of spurious relation is the full correlation
