Research methods and descriptive statistics
Test 2
Unit 12
The task of a scientist is to find out whether a causal statement is true (hypothesis testing).
Based on our testing we may reject a hypothesis or not, but we can never prove a causal statement.
Hypotheses can only be tested if they are precise.
Types of relationships:
,Deterministic means there are no exceptions to the line we expect to find. This is not what we often
find. Only if the expected relationship is deterministic, we can sometimes reject an expectation with
one single observation.
However in social sciences all expected relationships are probabilistic.
Why would we only expect probabilistic causality:
- Measurement errors; it is very difficult to have perfect measurements, so when we look at
reality we expect that things are a bit different from what we’d expect. Because we do not
observe correctly. This is why the dots aren’t always on the line.
- Parsimoneous models: we aim for simple models, so we’ve omitted variables/leave variables out
that we know are affecting our consequence. But we like to have a simple picture of the world,
so we ignore that. That’s why the dots aren’t always on the line.
The only way to test a causal relationship is by finding ‘variation’ in some way.
You can however come up wit a hypothesis without seeing differences or apply a causal statement to
a specific case when there is no variation. This is not really about testing, but more about applying.
you can find causality by:
- Within case analysis; comparing variation between observations in time of one individual.
Always think about the setting and if you can apply your hypothesis in a more general way
- Between/across case analysis; comparing people in one observation. We try to explain variance
In Y, by the variance in X
- Combined
Understanding causaul explanation.
Three implications of general causal statements:
1. Time order
The independent variable precedes the dependent variable.
2. Association
There is actually a correlation between the dependent and independent variable; the
consequence occurs less often if the cause is absent.
3. Non-spurious relationship
There isn’t a third variable that causes the correlation. If we do not check this, we can
establish that there is an correlation but we cannot establish that there is a causal
relationship.
two effects of a third variable, can be:
a. Explanation/confounding
The third variable has a relationship on both of the variables that causes these
variables to look causally related, but there is actually no relation.
, b. Specification/Interaction/Modification
The third variable can also have a relationship on the effect between the two
variables, this is the modification variable.
Test 2
Unit 12
The task of a scientist is to find out whether a causal statement is true (hypothesis testing).
Based on our testing we may reject a hypothesis or not, but we can never prove a causal statement.
Hypotheses can only be tested if they are precise.
Types of relationships:
,Deterministic means there are no exceptions to the line we expect to find. This is not what we often
find. Only if the expected relationship is deterministic, we can sometimes reject an expectation with
one single observation.
However in social sciences all expected relationships are probabilistic.
Why would we only expect probabilistic causality:
- Measurement errors; it is very difficult to have perfect measurements, so when we look at
reality we expect that things are a bit different from what we’d expect. Because we do not
observe correctly. This is why the dots aren’t always on the line.
- Parsimoneous models: we aim for simple models, so we’ve omitted variables/leave variables out
that we know are affecting our consequence. But we like to have a simple picture of the world,
so we ignore that. That’s why the dots aren’t always on the line.
The only way to test a causal relationship is by finding ‘variation’ in some way.
You can however come up wit a hypothesis without seeing differences or apply a causal statement to
a specific case when there is no variation. This is not really about testing, but more about applying.
you can find causality by:
- Within case analysis; comparing variation between observations in time of one individual.
Always think about the setting and if you can apply your hypothesis in a more general way
- Between/across case analysis; comparing people in one observation. We try to explain variance
In Y, by the variance in X
- Combined
Understanding causaul explanation.
Three implications of general causal statements:
1. Time order
The independent variable precedes the dependent variable.
2. Association
There is actually a correlation between the dependent and independent variable; the
consequence occurs less often if the cause is absent.
3. Non-spurious relationship
There isn’t a third variable that causes the correlation. If we do not check this, we can
establish that there is an correlation but we cannot establish that there is a causal
relationship.
two effects of a third variable, can be:
a. Explanation/confounding
The third variable has a relationship on both of the variables that causes these
variables to look causally related, but there is actually no relation.
, b. Specification/Interaction/Modification
The third variable can also have a relationship on the effect between the two
variables, this is the modification variable.
