Empirical Methods

Question 1

What is internal validity ?

Answer 1

If the quality of the research design, and the quality of the data are good

Question 2

What is external validity?

Answer 2

Can the results be generalized beyond the particular study (other populations, times)

Question 3

What is construct validity?

Answer 3

Do the mesures used capture the construct that is the focus of the study

Question 4

What’s the reliability (fiabilité)

Answer 4

Does the measurement produce stable results lead to identical results if the situation is unchanged

Question 5

What’s an economic experiment ?

Answer 5

We observe the behavior of real people who are motivated by real economic incentive. We observe the behavior under controlled conditions

Question 6

What is design?

Answer 6

It’s a specific combination of treatments to investigate a particular question

Question 7

What are the different design the principles ?

Answer 7

-payment
-control
-randomization
-deception
-realism and framing
-efficient design

Question 8

What is a partner design?

Answer 8

The same proposal and responder are matched over several periods

Question 9

What is a stranger design?

Answer 9

In every repetition, you randomly rematch the responders and the proposers

Question 10

What is the perfect stranger design?

Answer 10

Random rematch with a new proposer in every period

Question 11

What are the mesures of Central location to describe Data ?

Answer 11

Mean, median and mode

Question 12

What is standard deviation?

Answer 12

C’est l’écart type, il sert à visualiser si une valeur est extrême

Question 13

What is the problem about standard deviation and what is the remedy?

Answer 13

The problem is the comparison between standard deviation depends on the scaling, and the remedy is the coefficient of variation

Question 14

What’s the problem about the coefficient of variation and what is the remedy?

Answer 14

The problem is standard deviation is sensitive to outliers (valeurs aberrantes) and the solution is using the interquartile range (difference between the 25th and a 75th percentile)

Question 15

What are the mesures of dispersion to describe data?

Answer 15

Standard deviation, coefficient of variation and interquartile range

Question 16

How can we measure inequality in income distribution or concentration in the markets ?

Answer 16

Relative concentration : a small fraction of population earns/owns a large share of the pie
- lorenz curve
- Gini coefficient
Absolute concentration : a few companies control the markets or have a large market share
- numerical : Herfindahl index

Question 17

What is a Lorenz curve ?

Answer 17

It’s a cumulative share of people/data

Question 18

What the Gini coefficient ?

Answer 18

The Gini coefficient is a numerical measure for the distance of the Lorenz curve from the equal outcome (the 45 degree line)

Question 19

What is the Herfindahl index?

Answer 19

The Herflindahl index is an index who measuring market concentration. It’s widely applied in the competition law and antitrust policy.

Question 20

How to show two measures relate to each other ?

Answer 20

-correalation coefficient
-scatterplot
-linear regression

Question 21

What are we doing in descriptive statistics

Answer 21

It is a method to describe the data in our simple (example : female students have on average higher grades than male students).

Question 22

What is statistical inference ?

Answer 22

Statistical inference is investigate a subsample of the population and draw conclusions for the entire population

Question 23

What are the two types of error in statistical testing?

Answer 23

Alpha-error is rejected H0 when H0 is true.
Beta-error is fail to reject H0 when H1 is true

Question 24

What is a P value?

Answer 24

The p-value is the probability of observing an effect, at least as extreme as the one in your sample data, if the null hypothesis is true.

Question 25

What is a non-parametric test

Answer 25

A non-parametric tests does not require assumption about the distribution that underlie the observation (but still relies on the assumption of random sampling)

Question 26

When we will use a X^2-test and fisher’s exact

Answer 26

If both the dependent and implement variable are categorical, we use the Fischer’s exact (for small sample) or the X^2-test

Question 27

What’s the difference between X^2-test or fisher’s exact?

Answer 27

Fisher’s exact gives exact p-value but it’s computionally demanding for large n.
X^2 is an approximation that saves computational ressources (and the researchers times) it’s requires an expected frequency of more than 5observations in each cell.

Question 28

In which case we will use parametrics methods (t-test, regression)

Answer 28

When the dependent variable is normally distributed.

Question 29

In which case we will use non-parametric method (Wilcoxon tests) ?

Answer 29

When distribution of dependent variable is unclear.

Question 30

What is the confidence interval?

Answer 30

It’s an interval that contain the true value is there a probability (usually 95%)

Question 31

How we can generate causal evidence for our experiment ?

Answer 31

Causal evidence can be generated by controlled experiments, controlled experiment are not possible they are alternatives like :
-instrumental variable
-regression discontinuity

Question 32

What is regression discontinuty design (RDD)

Answer 32

RDD is used to estimate the causal effect intervention or treatments based on assignment to a program according to eligibility threshold.

Question 33

What is an instrumental variable?

Answer 33

There is a variable called instrumental which affects the X treatment but doesn’t correlate with the other unobserved variables

Question 34

What should be the characteristic of the instrumental variable to be valid?

Answer 34

-the instrumental variable should be randomly assigned
-the instrumental variant needs to satisfy the exclusion restriction : the instrument is not allowed to have an effect on inobservable variable.
-the instrument needs to be coordinated with the variable of interest X.