正在学习
6.8 Properties of Model Selection Procedures
6.8 Properties of Model Selection Procedures
6.9 Risk for Model Selection Methods: Monte Carlo Simulations
6.10 Conclusion
6.11 Appendix: Derivation of Information Criteria
II Forecast Methods
7 Univariate Linear Prediction Models
7.1 ARMA Models as Approximations
7.2 Estimation and Lag Selection for ARMA Models
7.3 Forecasting with ARMA Models
7.4 Deterministic and Seasonal Components
7.5 Exponential Smoothing and Unobserved Components
7.6 Conclusion
8 Univariate Nonlinear Prediction Models
8.1 Threshold Autoregressive Models
8.2 Smooth Transition Autoregressive Models
8.3 Regime Switching Models
8.4 Testing for Nonlinearity
8.5 Forecasting with Nonlinear Univariate Models
8.6 Conclusion
9 Vector Autoregressions
9.1 Specification of Vector Autoregressions
9.2 Classical Estimation of VARs
9.3 Bayesian VARs
9.4 DSGE Models
9.5 Conditional Forecasts
9.6 Empirical Example
9.7 Conclusion
10 Forecasting in a Data-Rich Environment
10.1 Forecasting with Factor Models
10.2 Estimation of Factors
10.3 Determining the Number of Common Factors
10.4 Practical Issues Arising with Factor Models
10.5 Empirical Evidence
10.6 Forecasting with Panel Data
10.7 Conclusion
11 Nonparametric Forecasting Methods
11.1 Kernel Estimation of Forecasting Models
11.2 Estimation of Sieve Models
11.3 Boosted Regression Trees
11.4 Conclusion
12 Binary Forecasts
12.1 Point and Probability Forecasts for Binary Outcomes
12.2 Density Forecasts for Binary Outcomes
12.3 Constructing Point Forecasts for Binary Outcomes
练习题
Which of the following is NOT a property of model selection procedures?
A. Consistency
B. Efficiency
C. Overfitting
D. Parsimony
What are the key aspects of risk for model selection methods in Monte Carlo simulations? (Select all that apply)
A. Estimation error
B. Model misspecification
C. Data sampling variability
D. Computational complexity
The conclusion of model selection emphasizes that no single method is universally best for all situations.
The derivation of information criteria is based on minimizing the expected value of the ___.
Explain the role of ARMA models as approximations in time series analysis.
Which criterion is commonly used for lag selection in ARMA models?
A. R-squared
B. Akaike Information Criterion (AIC)
C. Mean Absolute Error (MAE)
D. Coefficient of Determination
What are the key steps in forecasting with ARMA models? (Select all that apply)
A. Model identification
B. Parameter estimation
C. Diagnostic checking
D. Data transformation
Deterministic components in ARMA models include trends and seasonal effects, which are modeled separately from stochastic components.
Exponential smoothing assigns exponentially ___ weights to past observations.
Summarize the conclusion of univariate linear prediction models.
Threshold Autoregressive (TAR) models are characterized by:
A. A single regime with constant coefficients
B. Multiple regimes with different coefficient values
C. No autoregressive components
D. Only deterministic components
Which of the following are true about Smooth Transition Autoregressive (STAR) models? (Select all that apply)
A. They allow for gradual transitions between regimes
B. They use a logistic function for transition
C. They are a special case of TAR models
D. They assume abrupt regime changes
Regime switching models assume that the regime-generating process is observable.
Testing for nonlinearity often involves examining whether the ___ of residuals changes over time.
Explain the challenges of forecasting with nonlinear univariate models.
The conclusion of univariate nonlinear prediction models emphasizes that:
A. Linear models are always superior
B. Nonlinear models are universally better
C. Model choice depends on data characteristics
D. Nonlinearity is irrelevant for forecasting
Which of the following are key considerations in specifying Vector Autoregressions (VARs)? (Select all that apply)
A. Variable ordering
B. Lag length selection
C. Deterministic terms
D. Model estimation method
Classical estimation of VARs assumes that the error terms are normally distributed.
Bayesian VARs incorporate prior information through ___.
Describe the role of DSGE models in forecasting.
Conditional forecasts in VARs require:
A. Fixing some future values of variables
B. Using only past data
C. Ignoring deterministic terms
D. Assuming normality of errors
What are common challenges in empirical examples of VARs? (Select all that apply)
A. Model misspecification
B. Overfitting
C. Data non-stationarity
D. Computational efficiency
The conclusion of VARs states that they are always the best choice for multivariate forecasting.
Forecasting with factor models relies on extracting ___ factors from high-dimensional data.
Explain the estimation of factors in factor models.
Which criterion is commonly used to determine the number of common factors in factor models?
A. R-squared
B. Eigenvalue ratio
C. Akaike Information Criterion (AIC)
D. Bayesian Information Criterion (BIC)
What are practical issues in factor models? (Select all that apply)
A. Rotation invariance of factors
B. Sensitivity to outliers
C. Choice of estimation method
D. Interpretability of factors
Empirical evidence suggests that factor models outperform univariate models in all forecasting contexts.
Forecasting with panel data combines ___ and time series dimensions to improve accuracy.
Summarize the conclusion of forecasting in a data-rich environment.
Kernel estimation of forecasting models is an example of:
A. Parametric estimation
B. Nonparametric estimation
C. Bayesian estimation
D. Classical estimation
What are advantages of sieve models? (Select all that apply)
A. Flexibility in functional form
B. Computational simplicity
C. Avoidance of overfitting
D. Theoretical guarantees
Boosted regression trees are immune to overfitting.
The conclusion of nonparametric forecasting methods emphasizes the importance of balancing ___ and computational feasibility.
Explain the difference between point and probability forecasts for binary outcomes.
Density forecasts for binary outcomes specify:
A. The probability of each outcome
B. The full distribution of possible outcomes
C. Only the most likely outcome
D. The variance of outcomes
What are methods for constructing point forecasts for binary outcomes? (Select all that apply)
A. Maximum likelihood estimation
B. Logit/probit models
C. Linear probability model
D. Kernel regression
When evaluating the properties of model selection procedures, which of the following is NOT a key consideration when comparing different methods?
A. Consistency of the estimator
B. Computational complexity of the method
C. The model's ability to forecast deterministic components
D. Risk of overfitting
Monte Carlo simulations are used to assess the risk of model selection methods. Which of the following are valid reasons for using Monte Carlo simulations in this context?
A. To estimate the sampling distribution of a statistic
B. To compare the performance of different model selection criteria
C. To determine the optimal number of lags in an ARMA model
D. To evaluate the impact of model misspecification on forecast accuracy
Forecasting with ARMA models requires selecting the appropriate number of lags, and this selection process can be informed by the properties of model selection procedures discussed in Section 6.8.
In classical estimation of VARs, the choice of estimator often depends on the ___ of the model and the desired properties of the estimator, such as consistency and efficiency.
Explain how the properties of model selection procedures can influence the choice of factors in forecasting with factor models.
登录后解锁笔记、知识点解析、AI 问答
立即登录