Parametric Flashcards
What is the p-value?
It’s a value helping us to determine whether a variable statistically significant. The p-value is the probability of committing a Type 1 error (this is a false positive).
We have to set a threshold for the accepted p-value.
We want the p-value to be as low as possible.
What is R^2?
R^2 tells us how much of the variability of the cost(CAPEX feks) can be explained by the model.
We want R^2 to be as close to 1 as possible.
What does it mean that it is crossectional?
Several entities (in our case projects) observerd at a particular point in time
what is epsilon_i?
Residual, the remaining cost that cannot be explained by our choice of cost drivers.
Is there any restirictions on the R.H.S variables?
No, they can be raised in any power. However, the beta coefficients must be linear. Otherwise we cannot use OLS
How can we find beta_1 in a univariate crosssectional OLS regression??
p_xy(sigma_y/sigma_x) = correation_xy(stddev_y/stddev_x)
How can we find beta_0 in a univariate crosssectional OLS regression?
beta_0=mean_y-mean_x*beta_1
What is the mos important assumption in OLS regression?
Population orthogonality condition: the cost driver is not correlated/effected by the residual.
However we can never know for sure if the condition is fulfilled.
What could be the reason that the assumption population orthogonality condition is violated?
Endogeneity issues
What are three types of endogeneity issues that can violate the population orthogonality condition?
- Omitted variable bias
- Systematic measurement error
- Errors in specification of causality
What is errors in specification of causality?
We want X to effect Y, if it is the other way around or that X and Y effects each other, or if they’re effected by another variable, then the have errors in specification of causality
What are systematic measurement errors?
The measurements are done wrongly so that the wrong correlation is shown. E.g women overshare depression, while men undershare
What is the omitted variable bias?
If a variable is not included in the regression and the excluded variable both exhibits an effect on the y-variable and is correlated with the included x-variable, then we will have omitted variable bias
What is a solution to the omitted variable bias?
Just add the missing variable. (but then we’ll get a multivariate regression model)
How do you find beta in a multivariate regression model?
beta = (x^Tx)^-1 * x^Ty
