Python for Data Analysis: Data Wrangling with Pandas, NumPy, and IPython 1st Edition

End of 2013 Kindle Update--> Many ebooks (not just Kindle) have problems with math formulas in LaTex. Others (like this) have code or pseudocode, and lots of tables, which are problematic at times. IF you get this book for Kindle in 2014 or late 13, you are in for a treat: not only the online goodies, but the entire ebook itself has been extensively revised for Kindle, including code and tables. They are outstanding!Our previous Kindle edition wasn't awful, but this is just awesome now. If you're tired of R glitches and complexity, consider the many new (and FREE) features Wes details in this fine text, especially tips for free libraries and APIs, including of course NumPy and others that used to require a lot more math than they do today. Wes even has many plug and plays, and if you have even beginning skills in any oop (Java/C#), this will be easier than starting R from scratch. It has "nearly" the stats of R, and much more on all kinds of big data, not just research data. Highly recommended for my fellow kindlers. The native object- recursion in Python apis is alone worth this compared to R functional workarounds, even though I use both. Prior to this book, you'd be spending a LOTof time putting all this together visiting forums, libraries and APIs online.IMPORTANT NOTE on previous negative reviews: This new update not only fixes many issues with the tables for Kindle, but as you probably know if you're a Panda person, the online functional documentation for the library has been massively updated between late '12 and late '13. The author (the creator of the API) takes advantage of this with the kindle links. This makes this book a MUCH better reference than the last edition, part due to the new update and part to the value of the community work on functions, types and methods, which of course this author often leads. SO, even if you have an older version of print, many of the deficiencies (that frankly were not this author's fault!) are gone there too, because the links are still active and much better for configuring your code than before.This still isn't meant to be a "documentation" book, but, with the newly updated links, there are few programs you can't build now, including a LOT more detail on the functions themselves, with good keywords to augment the already fine examples and exercises here. Also, much less "heavy" from a programming view than most O'reilly tomes-- this author obviously understands beginners, and though this is not a how to learn Python book, it IS now a much better how to pick up DA, including pandas, numpy and other plug ins.

This was a much needed book. Kudos to Wes Mckinney for starting Pandas and then writing a book about using it. There is a lot of public documentation/videos on Pandas, but this one does a great job of introducing the many ways Pandas can be used.After going back & forth on developing my own 'Table' structure, I decided on Pandas. As a lot of people have commented on Pandas, the API's look straightforward and simple until you start trying to do something useful. My point is, the experience gets a lot better and this book goes a long way towards that goal. The author does try to showcase how he would use Python (in general) and that can be useful for improving python skills.There are also some comments and explanation about speed/performance, with some discussion on numpy so it's useful primer, especially if you want to think about modifying numpy/pandas or developing your own high perf Python module.This book does assume that you are not a Python beginner but chances are you are not, if you are looking into Pandas. It is part-cookbook, part tutorial. I've read thru the book once, and then found focus chapters to be particularly useful when working on specific areas of my application.Beware that Pandas is still under active development, and the author does a pretty good job of pointing out potential issues, but there are some bugs that get fixed with multiple releases so some instructions may not work perfectly. [All examples are probably okay though, at the time of writing & I don't think there should be any concern with those].As far as organization goes, the book jumps around a little across topics but that cannot be avoided since they're so interrelated. No points off for that.-1 star for two reasons:a) Would have liked to have seen some more examples where people might want to use pandas in slightly different contexts although that will probably be fixed easily in the future.b) On the topic of performance, the author has clearly shown that pandas is high performance when it comes to fairly large data, BUT it is typically slower when working with small quantities that are better handled in list-of-list routines. I profiled a few of the functions as I was converting my own date parsers & list-of-list/dict-of-list structures and found pandas slightly slower in many cases. The same is true of numpy as well, as is anything that incurs the overhead of C-function calls but some ideas about breakeven quantities would be nice. Perhaps this can be handled in the next edition too.Overall, I recommend the book highly and belongs on the bookshelf of any active python developer who deals with in-memory data. For applications in econometric & financial analysis, this book is essential!

This is an excellent book, assuming you like the author's approach to computation/data. This is in a sense also a review for pandas. I should emphasize I am NOT a programmer, in the proper sense. I am a (computational) physicist and have transitioned all of my data storage/analysis to pandas, for reasons I'll explain.A large portion of my work is "exploratory", where I try out many different ideas, hoping something sticks. I've wasted a large amount of time hacking away trying to piece together a somewhat complicated calculation on fairly abstract data sets, only to eventually lose track of what physics I'm trying to do because of how sloppy things get. Again, I am not a programmer! Computation is a tool to me, and time spent trying to make a tool work is time away from the actual job.Though I have only used it in earnest for a few months now, Pandas has increased my productivity tremendously. The organization/philosophy behind the program is amazing. Often (though less and less, thankfully) I find myself reverting to my old habits of working with a sloppy mixture of dicts,np.arrays, and classes, because I feel like I can do it faster/easier than setting it up in pandas. I am never right. Once I set up the problem in pandas, everything I could possibly want to do flows naturally.If you work with physical data and perform relatively complex calculations/transformations on it, I strongly recommend pandas and this book. Regarding the book, I will only say that by reading the author's (of the code and book!) perspective you quickly gain an appreciation for how powerful pandas can be.