Customer Reviews

Regression Models for Categorical and Limited Dependent Variables (Advanced Quantitative Techniques in the Social Sciences)

5 star:		(6)
4 star:		(2)
3 star:		(0)
2 star:		(0)
1 star:		(0)

 

 

This book is especially useful to start understanding topics like ordered probit, multinomial logit, negative binomial regression and zero-inflated count models. Although it starts with a chapter on the linear regression model, it should not be mistaken for an introductory text. I would certainly advise readers with limited background in regression models to start with other books, like the one of Wooldridge (Introductory Econometrics). The quality of this book must be that I've yet to see a book that explains these topics more intuitively. That is not to say it is easy or without mathematics, it's not. It just looks like the mathematics is only used for better comprehension, not to give you the full proof. Furthermore, while reading it you get the feeling that the author understands what you, as a researcher, are interested in. This allows him to focus on the topics of interest, like model selection and testing and interpretation of output. So although this is not a cookbook, it may well be the closest thing to it, especially in combination with his new book on applying these models in Stata. It is a pity that the author stops short of non-parametric models (next edition?).

Since I do statistical modeling in industry, I was looking for a good book on Logistic regression that would give me a deep understanding of the subject; one that also had wide coverage (Poison regression, Tobit models, ..etc.). I decided on J. Scott Long's book, after considering Applied Logistic Regression by Hosmer and Lemeshow, and Limited Dependent and Qualitative Variables in Econometrics by Maddala. I must say I am very pleased with my choice. The topics are very clear, and the math is used as an aid to understanding, and you don't get bogged down in the math. It is a pleasure to read the book.

A nice review of MLE-based methods for categorical, limited, and ordinal dependent variables. Most social science data is best thought of as categorical, ordinal, etc., not interval, and so a readable treatment of one approach to the analysis of such data that does not rely on intervality assumptions is worthwhile.

The author has a very clear explanation of topics such as how MLE works, some numerical methods for maximizing, various tests associated with MLEs, etc., all written at a intermediate level. It's not too advanced so readers won't be driven off but also isn't a cookbook. Lots of nice examples throughout.

It's definitely in standard regression mode, which is not to say bad, just limited. It doesn't cover (or indeed discuss) topics such as categorical multivariate analysis, alternate loss functions for estimating categorical or ordinal regressions, including alternating least squares approaches or quantile regression, categorical or ordinal time series, or instrumental variables.... It's not the last word on the topic, but is certainly a solid first word.

This book is especially useful to start understanding topics like ordered probit, multinomial logit, negative binomial regression and zero-inflated count models. Although it starts with a chapter on the linear regression model, it should not be mistaken for an introductory text. I would certainly advise readers with limited background in regression models to start with other books, like the one of Wooldridge (Introductory Econometrics). The quality of this book must be that I've yet to see a book that explains these topics more intuitively. That is not to say it is easy or without mathematics, it's not. It just looks like the mathematics is only used for better comprehension, not to give you the full proof. Furthermore, while reading it you get the feeling that the author understands what you, as a researcher, are interested in. This allows him to focus on the topics of interest, like model selection and testing and interpretation of output. So although this is not a cookbook, it may well be the closest thing to it, especially in combination with his new book on applying these models in Stata (only available at Stata). It is a pity that the author stops short of non-parametric models (next edition?).

If you have to do statistical analysis where your dependent variable is a count, a dichotomy, categorical, or ordinal, and if you are not a grad student in a statistics department, this is a book for you. Long clearly illustrates the need for the different models, covers the essentials of each, and provides further references. Obviously, no book on such a range of topics could be complete - there are entire long books written on each of the chapters in this one. But this is a good place to start, and it is nice to have it all 'tied together' - this makes it easier to see the relationships among the models.

There are many books dealing with the types of models covered in this text. I find the explanations here, however, quite a bit clearer than those found elsewhere. The discussion of various ways of interpreting coefficients in each of the models is the most useful portion of the text. While many other texts touch on the difficulties of interpreting coefficients and perhaps offer an approach or two, the author thoroughly reviews multiple approaches common and unique to each of the models.

This book is very readable. The author correctly claims that this book has been designed following questions from his various students in understanding such regression models. There is clarity in each and every argument used together with alternative ways of interpretation and testing. Great book.

Book I am using for class. Is ok. This is not a teach yourself kind of book.