EC2C3

Question 1

Define correlation

Answer 1

If one variable changing is associated with the other variable changing on average

Question 2

3 reasons why a variable may be correlated

Answer 2

1. Casual effect of x on y
2. Casual effect of y on x
3. A confounder causes both x and y to change

Question 3

What is the treatment

Answer 3

variable we want to know the effect of

Question 4

What is the outcome

Answer 4

The variable effected by the treatment

Question 5

What is a mechanism

Answer 5

The reason the treatment affects outcome

Question 6

How do we determine the existence of a bias

Answer 6

If, on average, the estimate doesn’t estimate the desired casual effect

(includes effect of confounders)

Question 7

What is a positive bias

Answer 7

Confounder causes both outcome and treatment to change in the same direction

Meaning estimate of the casual effect is bigger (more positive) than the causal effect on average

(regardless of negative or positive direction)

Question 8

What is a negative bias

Answer 8

Confounder causes both the outcome and treatment to change in opposite directions

Meaning that the estimate of the causal effect is smaller (more negative) than the causal effect on average

Question 9

What is reverse causality

Answer 9

The outcome has a causal effect on the treatment

Can be a source of bias

Think about the direction of this bias due to reverse causality

Question 10

How do you determine the direction of bias due to reverse causality

Answer 10

If the only reason for correlation between treatment and the outcome is reverse causality, what sign of correlation would there be

Question 11

Simultaneity

Answer 11

Causal relationship and reverse causality

Situation when there is a causal relationship and also reverse causality