8 internautes sur 9 ont trouvé ce commentaire utile
- Publié sur Amazon.com
UPDATE: I'm knocking this down to three stars. After finishing a very tough statistics course, I started browsing this book again, and I realized that the section on "comparing two population proportions" makes no mention at all of having to meet a success-failure condition (which is mentioned for a single-proportion hypothesis test). In fact, the example regarding the drug Adderall does not meet this condition. You don't even have to calculate the pooled estimate to see this; we're told that 8 of 210 subjects receiving a placebo were "successes." This is an invalid test and the results are therefore meaningless. If the author sees this, I'd appreciate any feedback in the comments.
ORIGINAL REVIEW: I'm an old geezer who has found that he needs to dive back into statistics late in life. I never had any stats classes in college, though I did have calculus and one linear algebra class. Since then, I've learned a fair amount about statistics on my own, but that was a long time ago.
I originally got the same author's "Statistics Essentials for Dummies" book because I got the "Calculus Essentials for Dummies" book some months ago and was floored by how good it was. The essential statistics book let me down, largely because it moved too quickly, by which I mean the author sometimes waved a magic wand over a topic, threw out a formula, and that was that. I could plug in numbers but had no understanding.
This book suffers from that problem to a degree, but it provides a somewhat higher level of detail. A typical example is the standard deviation. Like many beginner texts, this one shows how to calculate it, but doesn't explain how it was developed, who came up with the idea for it, or why. The author does mention, at one point, that the distance from the point of inflection on a normal distribution curve to the mean is a single standard deviation, but that's as far as she goes. This is tantalizing; I would really love to understand *why* that is so. There is no time for that in a book of this scope. That means this is a beginning text for learning to apply statistical methods, not a book to explain the theory behind those methods.
Anyway, I plowed through the whole thing in the course of a headache-inducing weekend, and by that time, I had achieved my first milestone, being able to compute a simple linear regression. Because I moved so fast, there are doubtless some concepts that haven't stuck permanently, and I'll need to review. But the fact remains that, in a single weekend, I got through random variables, binomial and normal distributions, hypothesis testing, confidence intervals, z-tables, t-tables, and finally, linear regression. I was one happy geezer when I closed the book.
So it was worth the modest purchase price. And it presented a relatively painless introduction to the topic, without on the one hand diving into abstract mathematical proofs (there are no proofs of any kind), or on the other, leaping over large topics with no explanation whatever.
At least one other reviewer noted that the book is repetitive. That's true. For example, the author harps on the dangers of not investigating surveys and polls you read about in the press far too often. If you didn't already know not to do that, you probably wouldn't be interested in learning statistics anyway. These and other oft-repeated topics could have been replaced by more substantive examples and explanations.
But I can apply all of the principles and formulas described in the book--in fact, I did, several times, while reading it--and I've gotten a good bootstrap on learning statistics. My personal goal is to go far beyond this book, to learn about machine learning, and apply those concepts both in R and in programming. But to get there, I had to start with the basics. I think this was a good, but maybe not great, place to start. I have now bought Statistics for Dummies II, because it covers other topics, including logistic regression, that I'm interested in.