Week 3

Question 1

What is a truncated data sample?

Answer 1

A data sample where we do not observe any data above (or below) a certain threshold.

Question 2

Why should we not use the regular OLS method for truncated data?

Answer 2

It would be biased, since E(u_i | x_i) ≠ 0

Question 3

Derive the truncated model (you can assume normalit).

Answer 3

Question 4

What is a censored data sample? What is the difference with a truncated data sample?

Answer 4

In a censored data sample you can not observe values above (or below) a certain threshold, however we do know that these observations exist. In a truncated model we do not know that.

Question 5

Why can’t we use the regular OLS model on a censored data sample?

Answer 5

It is biased since E[u_i | x_i] ≠ 0

Question 6

Derive the Tobit Type 1 model (you may assume normality).

Answer 6

Question 7

Answer 7

Question 8

What is a corner solution response?

Answer 8

Something where we use the same model as for censored data (Tobit type 1), however we are not intrested y_i*, but in y_i (i.e. we are not intrested in the true value but the observed value, e.g. when requesting donations we do not care about how much someone wanted to give - this could be below 0 - but how much they actually gave).

This is thus a different interpretation. DIT MOET NOG WEL VERBETERD WORDEN DENK IK

Question 9

What is an assumption of the Tobit type 1 model?

Answer 9

Question 10

How to check if the assumption of the Tobit type 1 model holds?

Answer 10

Question 11

What is the main difference between the Heckit and Tobit type 1 model?

Answer 11