Lecture 5

Question 1

5 Reasons why your regression analysis mght not make sense

Answer 1

1. Multicollinearity
2. Heteroscadacity
3. Outliers
4. Reverse causality
5. Omitted variable bias

Question 2

Outlier

Answer 2

Data point that does not follow the general trend of the data, an extreme value

Question 3

How can you solve outliers?

Answer 3

1. What is the reason for the outlier? Can/ should you do something?
2. Check how sensitive your results are to the presence of the outlier
3. Throw the outlier out of the dataset when e.g. mismeasurement, error in the observation, data entry error.

Not because it is convenient to do so!!!

Question 4

Reverse causality

Answer 4

When y (also) causes x, not only x causing y

Question 5

How to check for reverse causality

Answer 5

• Timing of measurement (Sometimes, x is measured later than y)
• Statistical tests, to check wether changes in x precede changes in y

Question 6

Omitted variable bias

Answer 6

An excluded variable has some effect on your DV and it is correlated with at least one of your IVs

Question 7

How to solve omitted variable problem?

Answer 7

• Avoid simple regression models (with only one IV)

- Include variables that are most likely to be most important theoretically in explaining the DV

Question 8

Three different types of data

Answer 8

• Cross-sectional data (Observations at a given point in time)
• Time series data
• Panel data (Observations over a period of time)

Question 9

Robustness / sensitivity analysis

Answer 9

Determine how sensitive your results are to changes in the model. E.g. combinations of control variables, datasets, time frames.

Do the results remain? Results are robust

Question 10

Requirements for causation (4)

Answer 10

• Time lag
• Correlation
• Previous research
• Test model for other explanations (robustness)

Question 11

Possible problem when all variables are measured in one year

Answer 11

• Reversed causality

- Less of an argument for causality