Logistic Regression: A Self-Learning Text (Anglais) Relié – 1 décembre 2005
|Neuf à partir de||Occasion à partir de|
Il y a une édition plus récente de cet article:
Les clients ayant acheté cet article ont également acheté
Descriptions du produit
Aucun appareil Kindle n'est requis. Téléchargez l'une des applis Kindle gratuites et commencez à lire les livres Kindle sur votre smartphone, tablette ou ordinateur.
Pour obtenir l'appli gratuite, saisissez votre adresse e-mail ou numéro de téléphone mobile.
Détails sur le produit
En savoir plus sur l'auteur
Dans ce livre(En savoir plus)
Commentaires en ligne
Commentaires client les plus utiles sur Amazon.com (beta)
The format is 13 chapters, possibly representing the 13 or 14 weeks in a typical school term. Each chapter has a specific statement of teaching goals at the front, a summary outline of the course to date in the back, and a few pages of questions or exercises with answers. There appear to be sample data sets available, formatted for popular stats packages, but I did not figure out how they are made available. Within the main text of each chapter, every page reads like a blackboard lecture: equations on the left and narration on the right. The presentation uses a minimum of math, just a little algebra and exponentials in a few specific forms.
For the aspiring tool-user, this book may be worth a semester's tuition. I can fault it only for an annoying habit of writing out in words equations that appear on the same page ("e raised to the power of the sum of products ... ").
This book is NOT meant for people truly interested in the theory or practice of the exact computations. For example, its use of probability scarely mentions joint or conditional distributions. As a result, some of its formulas (e.g. p.48) come across as rote memorization, instead of natural expressions of the laws of probability. Lacking joint probability, the covariance matrix can not have meaning. It is just something produced, somehow, by an oracular computer program.
The repeated phrase, "according to statisticians ..." makes it very clear that statisticians are a breed distinct from intended audience. What they do is quite alien, but somehow, sometimes leaves the student with formulas to grind through.
Before you buy this book, be very clear about what you expect from it. Beginning students may get a lot from it. Readers already familiar with probability and some stats are likely to be disappointed.
You might wonder: what is LR good for? The answer: when you want to assess a dichotomous outcome on the basis of any kind of predictors. For example, to predict disease occurrence (0/1) on the basis of gender, behaviors, income, etc. Or to predict a behavior (0/1) on the basis of psychological scores, demographics, etc.
The book follows a "lecture plus commentary" style, where a primary didactic text is heavily annotated with sidebar comments, summaries, reviews of the material, and quizzes with answers. Overall this is a good thing and makes the book extremely well-suited for self study. However, it also makes it extremely long and moderately tedious to read at times. Unlike many stats books, however, it actually is readable.
The mathematics are held to a high school level (i.e., algebra), so it is suitable for any applied researcher or research consumer, although therefore probably not suitable for a professional statistician. Still, it is mathematically rigorous and requires to reader to work through a large number of (simple) formulas, contingency tables, and the like.
One odd omission is R: the book covers procedures for SAS, SPSS, and Stata, but not R. The authors' website appears not to be updated since the 2nd edition, and also does not cover R. That is a puzzling lacuna given the growing popularity of R in general and especially in bioinformatics. However, it is not a crucial flaw, since LR in R is not difficult and there are many examples online.
In summary: if you're an applied researcher in medicine, public health, psychology, etc., and want to learn about LR, get it.