Fitting Statistical Distributions: The Generalized Lambda Distribution and Generalized Bootstrap Methods

Rating:
Summary: What the "Doctor Ordered".
Review: I do a great deal of number mashing especially for time series data. Although a Gaussian assumption is typical for a variety of reasons the really interesting stuff (IMHO) is where K.F.G.'s theory drops the ball. The GLD and its variations is very flexible in 'fitting' distributions. The exposition in this book has sufficient clarity as well as numerous examples. It is definitely slanted towards applications although there is sufficient theory that a strict empiricist can confuse themselves if they so choose. The theory is presented in a manner that I was able to 'translate' it into an EXCEL/VB application. The examples were sufficient to provide checks on the translation. Some code is also included in the book but it is for an application I do not use. Also, there are extensive tables of 'nominal' lambda values given various moments and/or quantile representations. ... In short, IMHO, the GLD should be in every numerical toolbox and the book should be on one's shelf. Good Stuff!

Rating:
Summary: Good content; difficult to extract from theorems and lemmas
Review: In its totality, this book has some very good content. The GLD is a great tool for practitioners. However, the summarization of materials is very poor and the authors lack applied knowledge. For the GLD, there are several unnecessary pages of lemmas, theorems, and proofs. These might be of interest to the mathematics researcher but to no one else.

Unfortunately, the valuable information contained in these proofs are not well summarized. After implementing a solver for the GLD lambdas based on 4 input moments, I had to sift back through the text to find constraints on lambdas and on the independent variable. They provide extensive proofs for where the GLD function exists (who cares??) and later some proofs on where the probability function is non-negative (more to the point, but isn't this implied from the start?). Now I have 2 sets of constraints and have to figure out where to apply each.

The authors also demonstrate a clear lack of knowledge on empirical applications by explicitly rejecting the use of non-biased moment estimators in favor of biased moment estimators. They actually state that the first four non-biased moments (mean, variance, skew, and kurtosis) "would have some ns replaced by n-1."

3 stars because I cannot find GLD information in a comprehensive format beyond this book (the authors are ground breakers).

Rating:
Summary: typos
Review: This is a review of the sample page, which misspells "bootstrap" as "bootstap" and does not predispose me to buy the book!

Rating:
Summary: typos
Review: This is a review of the sample page, which misspells "bootstrap" as "bootstap" and does not predispose me to buy the book!