# Introductory multivariate statistics reference for beginners

I am from computer science department doing research in data mining and image mining. I remember the last course about stat was introductory to statistics and probability in general. Now I have this course in master degree Multivariate statistical analysis and it's only for 1 month and it's very difficult to understand a lot of concepts. I need an easy to follow book about the subject of Multivariate statistical analysis for beginner!

• What are the topics of your course? Does the course instructor recommend a book? – Momo Mar 11 '15 at 11:28
• The course instructor book is in Chinese! So i need English easy to follow book. I visited the library and i saw Richard Johnson Applied on multivariate analysis and it was very hard, and i saw Anderson introduction to Multivariate statistics and i could not understand the first page! – Mohammad Ahmed Mar 11 '15 at 11:41
• Your question does not contain enough information. Do you need a conceptual or a more mathematical introduction? What should be covered in the book? Based on my psychological background I'd probably recommend either the book from Tabachnik & Fidell or Hair et al.. Especially the last one is well written and easily understandable. However, they do not cover anything about data mining. – phx Mar 11 '15 at 11:48
• Have you searched for [references] machine learning on this site? There are already multiple answers for your question. – Tim Mar 11 '15 at 11:49
• I would like the book to be related to data mining and machine learning in general, but looks like those books are advance on me. With that being said i will be happy to have introductory book at least to understand something! i don't understand what is a conceptual or a more mathematical introduction? I am programmer so science words are hard for me :( – Mohammad Ahmed Mar 11 '15 at 11:51

Judging by your question, your program sounds similar to the many accelerated MS degrees with 6-8 week courses on each subject. I would recommend to go for statistics or econometrics texts for executive MBAs to survive. The EMBA level texts are easy to follow and do not expect strong math background. e.g. Wharton's EMBA program has Stat 613 as core course, and it uses Stine and Foster's text.

It's good to set the expectations right though. It's impossible to learn statistics without at least calculus and linear algebra, so if your definition of "easy" is without these two skills, then you're not going to learn anything useful in one month, but it's Ok. It's just the nature of these degrees, you only need to get an exposure to the field, i.e. very similar to EMBA objectives.

For programmers I'd recommend fun books such as R by Example in Springer's Use R! series. I read it while learning R already knowing statistics, but think that it can be used to learn both R and statistics. R is rather interesting language for programmers, it's based loosely on functional programming (FP) paradigm. That's why if your programmer friend is a hardcore programmer, he must know stuff like Haskell or Scala, and will feel comfortable picking a new FP language, especially because FP is fashionable again these days.

Another title in the same series is An Introduction to Applied Multivariate Analysis with R. If your friend is in good school, he will probably have SpringerLink access through his library, i.e. free PDF download of a book.

• "It's impossible to learn statistics without at least calculus and linear algebra" Across many disciplines (no names!) the majority of practitioners of statistics have little or no background in those subjects, or at least that they can recall. Naturally it limits their understanding of the foundations, but equally much publishable work using statistics comes from such researchers. But I agree that "easy to follow" is difficult to comment on. I've not met anyone who wanted a book difficult to follow. – Nick Cox Mar 11 '15 at 13:08
• Okay thanks Nick, i mean by easy to follow not for advance but for beginner (programmer not mathematician student). – Mohammad Ahmed Mar 11 '15 at 13:17
• Let me see your response Aksakal and i will review it. Yeas actually i am for 2 years not rush master, but the course is for one month !! – Mohammad Ahmed Mar 11 '15 at 13:19
• @MohammadAhmed, you don't need to be math student to know calculus or linear algebra, it's a requirement in many quant majors, such as physics or engineering. Programmers are a different kind, they're sort of engineers, yet their math background is more like liberal arts. That's why you better look for texts not for engineers, but for "business" or liberal arts folks. – Aksakal Mar 11 '15 at 13:20
• @MohammadAhmed, "for Dummies", really? – Aksakal Mar 11 '15 at 17:53

Multivariate statistics is very broad, but given your background I'd recommend "Introduction to Statistical Learning" http://www-bcf.usc.edu/~gareth/ISL/ It's a really nice intro book to modern multivariate techniques that are popular in data mining and machine learning. You can get the PDF for free.

• Thank you so much, but i need one which is easy to follow not advanced or hardcore. I visited the library and i saw Richard Johnson Applied on multivariate analysis and it was very hard, and i saw Anderson introduction to Multivariate statistics and i could not understand the first page!! So as you see i am fresh to statistics and this class is a must for me to do. – Mohammad Ahmed Mar 11 '15 at 11:35
• The first one aims at being a gentle introduction. It won't get much easier than that. – Momo Mar 11 '15 at 12:21
• You mean Richard Johnson Applied on multivariate analysis? So what i some concept are hard! Do you have prerequisite book for it? – Mohammad Ahmed Mar 11 '15 at 12:37
• No, I mean the book in my answer! – Momo Mar 11 '15 at 15:09