具体描述
Recent decades have witnessed explosive growth in new and powerful tools for timeseries analysis. These innovations have overturned older approaches to forecasting, macroeconomic policy analysis, the study of productivity and long-run economic growth, and the trading of financial assets. Familiarity with these new tools on time series is an essential skill for statisticians, econometricians, and applied researchers.
Introduction to Time Series Using Stata provides a step-by-step guide to essential timeseries techniques from the incredibly simple to the quite complex and, at the same time, demonstrates how these techniques can be applied in the Stata statistical package. The emphasis is on an understanding of the intuition underlying theoretical innovations and an ability to apply them. Real-world examples illustrate the application of each concept as it is introduced, and care is taken to highlight the pitfalls, as well as the power, of each new tool.
作者简介
Sean Becketti is a financial industry veteran with three decades of experience in academics, government, and private industry. Over the last two decades, Becketti has led proprietary research teams at several leading financial firms, responsible for the models underlying the valuation, hedging, and relative value analysis of some of the largest fixed-income portfolios in the world.
目录信息
List of figures
Preface
Acknowledgments
1 Just enough Stata
1.1 Getting started
1.1.1 Action first, explanation later
1.1.2 Now some explanation
1.1.3 Navigating the interface
1.1.4 The gestalt of Stata
1.1.5 The parts of Stata speech
1.2 All about data
1.3 Looking at data
1.4 Statistics
1.4.1 Basics
1.4.2 Estimation
1.5 Odds and ends
1.6 Making a date
1.6.1 How to look good
1.6.2 Transformers
1.7 Typing dates and date variables
1.8 Looking ahead
2 Just enough statistics
2.1 Random variables and their moments
2.2 Hypothesis tests
2.3 Linear regression
2.3.1 Ordinary least squares
2.3.2 Instrumental variables
2.3.3 FGLS
2.4 Multiple-equation models
2.5 Time series
2.5.1 White noise, autocorrelation, and stationarity
2.5.2 ARMA models
3 Filtering time-series data
3.1 Preparing to analyze a time series
3.1.1 Questions for all types of data
How are the variables defined?
What is the relationship between the data and the phenomenon of interest?
Who compiled the data?
What processes generated the data?
3.1.2 Questions specifically for time-series data
What is the frequency of measurement?
Are the data seasonally adjusted?
Are the data revised?
3.2 The four components of a time series
Trend
Cycle
Seasonal
3.3 Some simple filters
3.3.1 Smoothing a trend
3.3.2 Smoothing a cycle
3.3.3 Smoothing a seasonal pattern
3.3.4 Smoothing real data
3.4 Additional filters
3.4.1 ma: Weighted moving averages
3.4.2 EWMAs
exponential: EWMAs
dexponential: Double-exponential moving averages
3.4.3 Holt–Winters smoothers
hwinters: Holt–Winters smoothers without a seasonal component
shwinters: Holt–Winters smoothers including a seasonal component
3.5 Points to remember
4 A first pass at forecasting
4.1 Forecast fundamentals
4.1.1 Types of forecasts
4.1.2 Measuring the quality of a forecast
4.1.3 Elements of a forecast
4.2 Filters that forecast
4.2.1 Forecasts based on EWMAs
4.2.2 Forecasting a trending series with a seasonal component
4.3 Points to remember
4.4 Looking ahead
5 Autocorrelated disturbances
5.1 Autocorrelation
5.1.1 Example: Mortgage rates
5.2 Regression models with autocorrelated disturbances
5.2.1 First-order autocorrelation
5.2.2 Example: Mortgage rates (cont.)
5.3 Testing for autocorrelation
5.3.1 Other tests
5.4 Estimation with first-order autocorrelated data
5.4.1 Model 1: Strictly exogenous regressors and autocorrelated disturbances
The OLS strategy
The transformation strategy
The FGLS strategy
Comparison of estimates of model
5.4.2 Model 2: A lagged dependent variable and i.i.d. errors
5.4.3 Model 3: A lagged dependent variable with AR(1) errors
The transformation strategy
The IV strategy
5.5 Estimating the mortgage rate equation
5.6 Points to remember
6 Univariate time-series models
6.1 The general linear process
6.2 Lag polynomials: Notation or prestidigitation?
6.3 The ARMA model
6.4 Stationarity and invertibility
6.5 What can ARMA models do?
6.6 Points to remember
6.7 Looking ahead
7 Modeling a real-world time series
7.1 Getting ready to model a time series
7.2 The Box–Jenkins approach
7.3 Specifying an ARMA model
7.3.1 Step 1: Induce stationarity (ARMA becomes ARIMA)
7.3.2 Step 2: Mind your p’s and q’s
7.4 Estimation
7.5 Looking for trouble: Model diagnostic checking
7.5.1 Overfitting
7.5.2 Tests of the residuals
7.6 Forecasting with ARIMA models
7.7 Comparing forecasts
7.8 Points to remember
7.9 What have we learned so far?
7.10 Looking ahead
8 Time-varying volatility
8.1 Examples of time-varying volatility
8.2 ARCH: A model of time-varying volatility
8.3 Extensions to the ARCH model
8.3.1 GARCH: Limiting the order of the model
8.3.2 Other extensions
Asymmetric responses to “news”
Variations in volatility affect the mean of the observable series
Nonnormal errors
Odds and ends
8.4 Points to remember
9 Models of multiple time series
9.1 Vector autoregressions
9.1.1 Three types of VARs
9.2 A VAR of the U.S. macroeconomy
9.2.1 Using Stata to estimate a reduced-form VAR
9.2.2 Testing a VAR for stationarity
Other tests
9.2.3 Forecasting
Evaluating a VAR forecast
9.3 Who’s on first?
9.3.1 Cross correlations
9.3.2 Summarizing temporal relationships in a VAR
Granger causality
How to impose order
FEVDs
Using Stata to calculate IRFs and FEVDs
9.4 SVARs
9.4.1 Examples of a short-run SVAR
9.4.2 Examples of a long-run SVAR
9.5 Points to remember
9.6 Looking ahead
10 Models of nonstationary time series
10.1 Trends and unit roots
10.2 Testing for unit roots
10.3 Cointegration: Looking for a long-term relationship
10.4 Cointegrating relationships and VECMs
10.4.1 Deterministic components in the VECM
10.5 From intuition to VECM: An example
Step 1: Confirm the unit root
Step 2: Identify the number of lags
Step 3: Identify the number of cointegrating relationships
Step 4: Fit a VECM
Step 5: Test for stability and white-noise residuals
Step 6: Review the model implications for reasonableness
10.6 Points to remember
10.7 Looking ahead
11 Closing observations
11.1 Making sense of it all
11.2 What did we miss?
11.2.1 Advanced time-series topics
11.2.2 Additional Stata time-series features
Data management tools and utilities
Univariate models
Multivariate models
11.3 Farewell
References
Author index
Subject index
· · · · · · (收起)
读后感
用户评价
这本书的配套资源，尤其是那些代码示例，在我看来是衡量一本技术书籍成功与否的关键指标之一。如果只是纸上谈兵，那些公式和理论即便再完美，也难以转化为实际的生产力。我猜想，这本书很可能包含了一套完整的、可以立刻在Stata中运行的练习文件。这种“即学即练”的模式对于学习效率的提升是显而易见的。更进一步讲，优秀的教材不会止步于展示正确的代码，它还会展示“错误的尝试”以及如何调试。比如，当数据不满足模型假设时，Stata会返回什么样的错误信息？读者应该如何根据这些反馈信息来修正自己的分析步骤？这种对“调试过程”的重视，体现了作者对真实科研环境的深刻理解。如果这本书能像一位耐心且经验丰富的导师那样，引导读者走过这些“弯路”，那么它的价值将远远超过其定价。
从一个资深数据分析师的角度来看，我更关心这本书对于高级主题的处理深度。仅仅停留在经典的ARIMA框架是远远不够的，当前的时间序列分析领域早已拓展到了更复杂的模型，比如状态空间模型、非线性时间序列模型，甚至是涉及高频数据的处理方法。我希望这本书能够超越基础的教学目标，为有一定基础的读者提供一些“干货”。例如，在处理多变量时间序列时，VAR（向量自回归）模型的选择、滞后阶数的确定，以及脉冲响应函数（IRF）的解释，这些都是实操中经常遇到的难点。如果这本书能够提供细致的Stata代码片段，并对不同模型的优劣势进行横向比较，那它就不再仅仅是一本入门指南，而可以成为案头常备的参考手册。我特别期待看到它如何处理时间序列中的异常值和结构性断点问题，因为这些“脏数据”往往是导致模型失效的罪魁祸首。
这本关于时间序列分析的书，虽然我还没来得及细读，但从初步的浏览来看，它的体例和覆盖面确实令人眼前一亮。我特别欣赏它那种深入浅出的叙述方式，仿佛作者不是在写一本教科书，而是在和一个经验丰富的同事进行面对面的交流。对于初学者而言，很多时间序列的经典教材往往开篇就抛出复杂的数学公式，让人望而却步，但这本书似乎避开了这种陷阱，而是选择了一条更具实践性的路径。它很可能在介绍基本概念时，着重于直观的理解和实际案例的展示，而不是纯粹的理论推导。我期待它能在我搭建自己的计量模型时提供可靠的脚手架，尤其是在处理那些非平稳性数据时，我相信它会给出非常具体的Stata操作步骤和结果解读指南。这本书的价值，或许不在于它创造了多少新的理论，而在于它如何有效地将复杂的分析技术转化为工具箱里的实用方法，让使用者能够迅速上手，解决现实世界中的数据问题。这种注重“如何做”而不是仅仅“为什么”的倾向，对于需要快速产出分析报告的研究人员来说，无疑是极具吸引力的特性。
我不得不说，这本书的装帧和排版设计，给我的第一印象是非常专业且严谨的。清晰的字体和合理的行距，使得长时间阅读的疲劳感大大降低，这对于一本技术性较强的书籍来说至关重要。当然，更重要的是内容组织上的逻辑性。我注意到章节之间的过渡非常自然，似乎每增加一个知识点，都是建立在前一个知识点之上，形成了一个稳固的知识梯队。例如，在讲解自回归模型（AR）和移动平均模型（MA）时，作者很可能花费了大量篇幅来解释它们的内在联系和局限性，然后顺理成章地过渡到ARMA和ARIMA模型的构建过程。这种层层递进的教学结构，极大地减少了读者在学习路径上的迷茫感。我个人对于时间序列的“预处理”阶段尤其关注，这本书是否详尽地介绍了如何进行单位根检验、协整检验，以及最重要的——如何选择合适的模型结构？如果这些基础工作能被详尽地阐述，那么这本书的实用价值将大大提升，因为它深知，模型选取的成败往往取决于数据预处理的细致程度。
总而言之，我对这本聚焦于Stata的时间序列教材抱有极高的期望。它似乎不仅仅是一本关于理论的书，更像是一份详尽的操作手册，旨在弥合统计理论与实际软件应用之间的鸿沟。我特别关注它在讲解模型评估和选择时所采用的判断标准，比如AIC、BIC准则的应用细节，以及残差分析的严格性要求。在时间序列分析中，模型的稳健性至关重要，任何一个轻微的设定错误都可能导致灾难性的预测结果。因此，我希望作者在书中强调模型诊断的每一个细节，教导读者如何对模型结果保持批判性的审视态度，而不是盲目地相信软件输出的“最佳”结果。这本书如果能成功地培养出读者这种审慎的研究习惯，那么它将是一份无价的学习资产，能够帮助我们构建出更可靠、更具解释力的时间序列模型。
超级好的书
超级好的书
超级好的书
超级好的书
超级好的书
