It is very common in the natural sciences to have exact analytical results, but to lack the mathematical techniques to provide analytical solutions to the resulting equations. In many cases this is due to incomplete knowledge, so some future mathematician will come up with solutions that do not now exist. However, it is often the case that there do not exists closed-form solutions, or the problem is so large that the required calculations are infeasible. The latter is often the case with so-called complex systems---they are complex only in the sense that the solution space exceeds our capacity to calculate.
In such situations, the accepted research technique is to find approximate solutions for an appropriate range of model parameters. This book is devoted to providing techniques for this "computational" approach. The authors' preferred models are dynamic optimization models, and somewhat ironically, their presentation of the models (although not going beyond Bellman and Lucas-Stokey) is the more interesting part of the book. By contrast, their presentation of computational methods is elementary, basically describing a tool kit using the MatLab software environment. This is a serious error, because it leads the user away from useful computational techniques.
The book opens with techniques for solving linear equations and approximating roots to continuous functions. In fact, the user rarely needs to know such details, but rather should go to Mathematica or Maple software that can do a better job in 99% of the cases that a casual user will ever do knowing the few classical techniques used in this book. But the authors never mention any software except MatLab, which is good for some things but not very good for others.
The biggest gap in the books is its treatment of Markov processes, which are ubiquitous in models of choice and strategic interaction. Markov processes are classic examples of analytical models in which it is easy to write down the equilibrium and even the dynamics, but the equations are many orders of magnitude too numerous to solve in human dimensions of time and space. Moreover, in my estimation Markov models are much more important than dynamic optimization models, which presume much more information on the part of the decision-maker than is usually available (outside of an engineering context). Finanical economic is a mess in part because it makes assumptions that allow dynamic optimization to appear to provide useful solutions, but in fact more realistic behavioral models, taking seriously the information possessed by decision-makers, would be much more useful.
As an alternative to this book, I would look at Judd and Tesfatsion's Handbook of Computational Economic and the many references therein.