LDVS

In which of the following situations is it most appropriate to use a 0-1 dummy variable as the dependent variable in a model?

• Modeling the amount of dividend payout in dollars for each firm.

• Modeling why some firms choose to engage in stock splits while others do not.

• Modeling the total number of shares traded on the NASDAQ each day.

• Modeling the percentage change in a country’s GDP after a debt default.

Which of the following is NOT a reason why the Linear Probability Model (LPM) is often considered inadequate for modeling binary dependent variables?

• It can predict probabilities below 0 or above 1, which are not valid probabilities.

• Truncating predicted probabilities at 0 or 1 leads to too many observations at the extremes, which is unrealistic.

• The error term in the model cannot be assumed to follow a normal distribution.

• The LPM always produces perfectly homoscedastic errors, making standard errors too large.

Which of the following statements about the logit model is TRUE?

• The logit model transforms probabilities using the cumulative logistic distribution, ensuring predicted probabilities stay strictly between 0 and 1.

• The logit model is linear and can be estimated using ordinary least squares (OLS).

• The logit model guarantees that probabilities are exactly zero or one at the extreme values.

• The logit model is typically estimated using heteroscedasticity-robust OLS methods.

Which of the following statements about the logit and probit models is TRUE?

• The logit and probit models usually produce very different results because they use different distributions.

• The main historical reason for preferring the logit model was that it required evaluating a complex integral, making it slower to estimate.

• Today, the choice between logit and probit models is usually arbitrary because modern computational speeds make both equally practical.

• The logit and probit models are less preferred compared to the linear probability model because they are harder to interpret.

Why is it incorrect to interpret the coefficient β2​ in a logit or probit model as the percentage change in the probability when x2i​ increases by 1 unit?

• Because the coefficients are always measured in dollars, not percentages.

• Because the relationship between x2i​ and Pi​ is non-linear.

• Because logit and probit models do not estimate standard errors or t-ratios.

• Because hypothesis testing is not applicable in non-linear models.

Which of the following is a commonly reported goodness-of-fit measure for limited dependent variable models like logit or probit?

• Residual Sum of Squares (RSS)

• Pseudo-R2

• Mean Squared Error (MSE)

• Adjusted R2

Which of the following is an example of a decision involving more than two alternatives (not a simple binary choice)?

• Deciding whether to pay dividends or not.

• Choosing between listing on NYSE, NASDAQ, or AMEX.

• Predicting whether a country will default on its debt or not.

• Modeling whether a firm will issue bonds or not.

What is the key difference between truncated and censored variables?

• Truncation excludes data outside certain limits; censoring keeps all data but caps the values at a threshold.

• Censoring excludes data outside limits; truncation keeps all data but caps the values.

• Truncation applies only to binary variables; censoring applies only to continuous variables.

• Censoring and truncation are the same and can be used interchangeably.

What is the name of the standard approach used to estimate models with censored dependent variables?

• Logistic regression

• Probit regression

• Tobit analysis

• OLS (Ordinary Least Squares)