Seminars 1, 2 ,3 ,4 ,5, 6 ,7 Flashcards
What questions should we put ourselves first when we first observe a regression?
- About the variables: is there an economic theory on them?
- About missing variables: is there any bias from omitted variables?
- About linear specifications: is it advisable to have a linearizable form?
- About indivitual significance of coefficient: what are the interpretation?
- About multicollinearity: how can we avoid it?
What is the purpose of an economic theory in analysing variables?
An economic theory helps us predict the ways in which variables depend on each other.
Ex: higher education results in higher pay
What is ommited variable bias?
It occurs when a key independent variable is not included in the regression model, leading to an overestimation/understimation of that model.
What are some cases in which using the logaritmic function is a good ideea?
- When we want to interpret values in percentages
- When we want to solve for the asymmetry of the dependent variable
- Residuals have a “strongly” positively skewed distribution
- In the case in which we want to solve for heteroskedasticity
What is perfect multicolinearity?
It occurs when two identical variables are expressed in different measurement units
Example: dummy trap
How can you check for correlation between two variables in Eviews?
- Open the variables as a group
- Click View -> Covariance analysis
- Check only correlation box
What is near multicollinearity?
It occurs when there is a high correlation between the independent variables.
(>0.7 or 0.8)
How should you build an optimal regression model?
Include independent variables that are highly correlated with the dependent but not among themselves.
If you discover a pair of highly correlated independent variables, remove the one with the lowest correlation with the dependent one.
What are the thresholds for the p-value?
- p<1% (***)
- p>1%&<5% (**)
- p>5%&<10% (*)
What does the R squared coefficent show?
And the R squared adjusted coefficient
The R-squared coefficient shows the degree to which the variance of the dependent variable is explained by the independent variables.
R-squared increases with the number of variables included in the model.
ex: For R-squared=0.5 : 50% of the variance in crime rate is explained by the predictors included in the model, while the rest of 50% is explained by other predictors not included in the model
What is the Adjusted R-squared coefficient?
Adjusted R-squared adds a penalty to account for the problem of R-squared.
Meaning that its value goes down as more variables are added.