Important final subjects Flashcards
What is a t-test and how should I interpret it?
A t-test is Beta divided by standard error and measures how many standard errors the coefficient is away from zero (no effect). Therefore, the more extreme the t-value, the better.
What is a P-value and how should I interpret it?
The probability that the t-statistic of the coefficient would be as large as the one from the null-hypothesis; in other words; the probability that there is no effect. The smaller, the better.
What is the confidence interval and how should I interpret it?
The range in which the coefficient is likely to be with a certain percentage of confidence. If there is no 0 in the confidence interval, it is significant.
What is an F-test in multivariate OLS regression, and how should it be interpreted?
F test evaluates whether the model as a whole is statistically significant or the joint influence of the independent variables on the dependent variable. The bigger the better. An F test of 15 says that the model explains 15 times as much variance is what is left unexplained.
Describe how an omitted variable can lead to bias in estimators.
If the omitted variable is correlated with both the included variable and the dependent variable, the included variable describes more effect than is actually his!
How does a latent variable work in Tobit model?
A latent variable is unobserved and represents a condition or trait for the other variables that are measured. For example, if Y needs to be larger than 0, the latent variable says Y should only be observed if it is larger than 0, otherwise it gets the value zero.
How can we solve heteroscedasticity?
With robust standard errors. These are standard errors that are adjusted for heteroskedasticity, which is variance across the error terms of different observations.
What is the correct term when interpreting the economic significance?
IS ASSOCIATED WITH
When there is a log(independent variable), how should I interpret this?
A 1% increase in this variable, leads to Beta / 100 and then scaled to unit of measurement ….
If they ask, how to determine the statistical significance given the information with a t-test, how to answer?
For each parameter, we test H0: Beta = 0 or H1: Beta is not equal to zero.
This is a two-sided test.
For each test, we compare the absolute value (Beta/SE) with the critical value for the 10% level = 1.64.
If t>1.64, we reject the null hypothesis and the effect is significant.
