I started to read Handbook of Statistical Analysis & Data Mining - Applications by R. Nisbet, J. Elder, G. Miner and many guest authors just several days ago. Thus, this is not intended to be a review, this is my first impression and wish for your opinions.
I was looking forward to this book for a long time and after first brief pass through the handbook, I agree - it is fulfilling my anticipation completely. There are several good books focused on modern statistics sensu lato on the market. But this is the first really kind step-by-step-like tutorial for beginners. Not only that - it covers a lot of very diverse topics and applications providing the perfect evidence for nowadays (and future, of course) crucial importance of modern statistics in virtually any field. Moreover, many tutorials reach the professional level, revealing useful hints which are traditionally “secret know-how of experienced data miners and companies”. Several reasons why – in my opinion – this book is a great enterprise follow. I am looking forward to your opinions.
1. About 1/2 of pages are devoted to step-by-step tutorials (many other case studies and examples are on the attached DVD). Data and example outputs from statistical software are provided.
2. Virtually any field is covered (bioinformatics, risk management, CRM, scoring, finance modeling, industry and quality control, psychology, fraud detection, churn analysis, health…)
3. SAS-EM, SPSS-Clementine (Modeler) and STATISTICA Data Miner are used for example solving. That thorough opportunity to compare the 3 most common data miner software in one book is very outstanding.
4. The highest number of tutorials are performed in STATISTICA Data Miner (G. Miner´s affiliation is certainly partial catalyst of this). It is a great day for StatSoft, due to providing evidence, that STATISTICA Data Miner is able to deal with various very complex projects, at least as virtuously as more expensive competitors.
5. StatSoft released 90-day trial of SDM (on embedded DVD) for readers of the handbook. It is a good promotion of StatSoft, since trial of the SPSS and SAS data miners are very hardly available.
6. The book is not only a tutorial for “black box” modeling. Many statistical considerations, comparisons of a diverse method performance, data preparation and imputation purpose etc. are building an “analytic bridge” between the classical statistics and data mining. It removes a big piece of unbelief and uncertainty from data mining, which is still sometimes felt as a manipulation with reality and “data fishing”.
I will be goint through the handbook comprehensively and I believe that soon I will be able to add more specific findings. In the meantime, I am anticipating yours…
Jiri Kubasek, StatSoft CR