دانلود کتاب تجزیه و تحلیل سری های زمانی و کاربردهای آن با مثال های R

The basic layout of this edition of the text is similar to the last two editions. There are still seven chapters generally covering the same subject matter and data examples use R (R Core Team, 2024). There are four appendices, the first three cover the same material as the previous edition, and the fourth appendix is a complex number primer that replaces the R tutorial. The R tutorial has been moved online to dsstoffer.github.io/Rtoot. Warning: If loaded, the package dplyr may corrupt the base scripts filter and lag that we use often. In this case, to avoid problems, either detach the problem package detach(package:dplyr), or issue the commands filter=stats::filter and lag=stats::lag before analyzing time series data. We have used earlier versions of the text at both the undergraduate and graduate levels. Our experience is that an introductory undergraduate course can be accessible to students with a background in regression analysis and a first course in mathematical statistics (e.g., Spiegel et al., 2013) and may include Sects. 1.1–1.5, 2.1–2.3, the results and numerical parts of Sects. 3.1–3.9, and briefly the results and numerical parts of Sects. 4.1–4.5. The text Shumway and Stoffer (2019) covers these topics using a data science approach. At the advanced undergraduate or master’s level, where the students have a mathematical statistics background at at least the level of DeGroot and Schervish (2014), more detailed coverage of the same sections, with the inclusion of extra topics from Chap. 5 or Chap. 6 can be used as a one-semester course. Often, the extra topics are chosen by the students according to their interests. We have given a one-semester course on state space models using Chap. 6 by first briefly covering ARMA models in Chap. 3 and in Chap. 5. Finally, a two-semester graduate course for mathematics, statistics, and engineering students can be crafted by adding selected theoretical appendices, e.g., by including a discussion of how the central limit theorem extends to dependent data. For the graduate course, we should mention that we are striving for a broader but less rigorous level of coverage than that which is attained by Brockwell and Davis (2013). The package for the text, astsa, has been updated several times since the previous edition. Many of the data sets and scripts have been updated; in particular, Kalman filtering and smoothing and the EM algorithm have been updated. There are other new data sets and scripts that are used throughout this version. Information about the package can be found through various links at dsstoffer.github.io. To distinguish between the analysis of real data and of simulated data, graphics involving real data have a white background whereas simulated data have a gray background. For example, compare Fig. 1.1 with Fig. 1.9. In Chap. 1, a number of examples use data sets that have been updated to 2023 and there are also new examples. We have also included a discussion of random number generation (RNG) because it seems that students do not have much knowledge about its technical aspects. Because RNG can be based on difference equations, we felt that a course on time series is a good place for an introduction. In Chap. 2, some data sets have been updated or are different. For example, we introduce the El Ni˜no Southern Oscillation (ENSO) data set, which is nowused throughout the text. We have also included a discussion of classical structural modeling via smoothing. Chapter 3 has a similar layout as in the previous edition.We expanded bootstrapping AR models via a new script called ar.boot so that students do not have to do any coding (and ar.mcmc for Bayesian inference with details in Chap. 6). We found that some statistics students have a difficult time with the frequency domain analysis in Chap. 4. Consequently, in addition to a complex numbers primer (Appendix D), there are new examples that can also help introduce the topics. We also expanded the discussion of tapering. Finally, there is a new section (Sect. 4.11) at the end of the chapter that deals with discovering structural breaks in time series. It is an advanced topic that presents minimum description length and genetic algorithms. In Chap. 5, we removed the section on Box-Jenkins time domain transfer function modeling because Sects. 4.8, 4.9, and a fair amount of Chap. 7 are devoted to spectral domain lagged regression and transfer function modeling. Interested readers can find the Box-Jenkins method in various texts such asWei (2023, Chap. 14). Also, Chap. 5 now includes a discussion for tests of linearity. Chapter 6 on state space modeling has been expanded with new and faster Kalman filter, smoother, and EM algorithm scripts in astsa. There is now a discussion on the inclusion of inputs as well as an example of using the EM algorithm with constraints. The Bayesian analysis portion of the chapter has been expanded to include MCMC methods, particle filtering, effective sample size, and the fitting of stochastic volatility models (with feedback) using Bayesian and classical methods. Chapter 7 is essentially the same as the previous edition except that there is now a script to run the spectral envelope (see Sect. 7.9) called specenv. Examples using the spectral envelope on DNA sequences (categorical data) and real-valued time series are given in that section using the new script.