دانلود کتاب مقدمه‌ای بر نظریه مطالعات مشاهده‌ای

This book is a quick jog along one useful and attractive path offering brief but good views of the theory of causal inference in observational studies. It is tour of an island, with an invitation to stay, but if you choose to stay, then there is more to see. The path is attractive: I pick topics and methods that offer easy access to grand vistas. The path is useful: you could conduct excellent observational studies while staying on this path. The path quickly gets us out of base-camp into the hills; we quickly move beyond defining causal effects and making adjustments for observed covariates, to face the real challenges. That said, we will not reach all the points of interest, nor discuss all the useful methods. There are always ways to extract a bit more from data, to see a bit deeper into data, to design a better study, as well as many nearly equivalent ways to do any one thing; so, our path will not pass every site worth seeing. Chapters end with suggestions for Further Reading. Why does our path through the southeast corner of the island go up to summit A rather than to adjacent summit B? Here are some answers, and hopefully one will satisfy. (i) If you become obsessed with such questions, you will miss some fine views. (ii) The answers to such questions are often much more difficult than the climb to summit A, and the answers are always less attractive than the view from summit A. (iii) I have taken people up summit A and summit B, and there is a short scramble over boulders on the way up to the top of B—someone always sprains their ankle on the way up to B, needs urgent care, and is carried out of the woods screaming on a stretcher. This happens all the time in statistics courses. (iv) If we go up both summit A and summit B, then our quick tour will not have time for the very different views on the north and west of the island. The southeast corner is great, but there is so much more. (v) Trust me, not forever, just until we reach the discussion of Further Reading. The book stays focused on causal inference in observational studies, that is, on issues that arise because individuals were not assigned at random to treatment or control. In any large, complex empirical investigation, other issues will arise too, but I leave such issues to other books that focus on topics other than causal inference in observational studies. This book is an introduction to the theory of its subject. It is aimed at someone who wants to know how and why things are true, not someone who wants to be told what is true. As a consequence, basic knowledge of mathematical statistics, at the undergraduate level, is the prerequisite background. The proofs in the text and the problems at the end of chapters are aimed at someone who has taken an undergraduate course in mathematical statistics. In contrast, a proof in an appendix to a chapter is typically not difficult, but it may use slightly more technical tools. Appendices and sections or subsections with an asterisk may be skipped without encountering problems later in the book; however, appendices differ from sections with an asterisk. A section with an asterisk offers you more than you need to continue, something interesting but optional. By and large, an appendix demonstrates some specific technical fact needed in the text for which a reference is not available in the literature. An R package iTOS was created as a companion to this book. It is publicly available at CRAN. The package contains the data sets and reproduces selected analyses. Some of the problems ask you to do analyses using the iTOS package.