Michael Rutter

As as graduate student, I slowly began to realize that the data we collect for statistical analysis is not as simple or as clean as our introductory statistics texts would lead us to believe. One technique for dealing with data that, due to the sampling design, are correlated at many different levels, is a hierarchical or multilevel modeling approach. This book is an excellent introduction to this topic, with examples from a number of different subject areas. I find myself often turning to this book for inspiration on how to deal with a difficult dataset and for ways to help explain results to non-statisticians.