Data Analysis: A Model-Comparison Approach

by Charles M. Judd, Gary H. McClelland

College text

Genres: Textbooks, Nonfiction

Hardcover

First published January 1, 1989

Reviews

[Wej · Rating 4 out of 5]

Experimental scientists, as least in my days, were taught a recipe-style approach to statistics. For instance, if your study design is X then use test Y etc. This book introduces a more sensible, in my view, way of teaching inferential statistics. The authors focus on comparing models, which all work like linear regression. Once the basics of linear regression are taught, it is easy to see how the same approach can be used with multivariate statistical tests, and how looking for the best metric between models (say R2 or RMSE) is the best way to go. This approach won’t be new to economists, but in behavioural science (the authors work in psychology and use examples from this discipline) it is an epiphany.

Admittedly, there are other ways of analysing data than fitting and comparing models. Fitting data to a model can often yield significant results if the number of parameters is increased. However, outside of academic research simpler approaches (e.g. fast and frugal heuristics) or ML-based predictions can be even more useful.

I never took so long to read a book as with this one. One one hand, it is full of detailed analysis and the readers are given examples and step-by-step calculations of key metrics. But then the examples seemed dry and after a while reading this book felt like a chore. It did not help that this was a compulsory reading for my statistics class. This meant that each time I opened it, it felt like homework.

[Pauline · Rating 5 out of 5]

Super clear, very interesting, and did what it's supposed to do: taught stats in an easy a way as possible.