The logic of survey analysis.

by Morris Rosenberg

1968. Second Printing. 283 pages. Blue dust jacket over blue cloth covered boards with gilt. Pages have mild tanning and foxing throughout, moderate at end-papers, paste-downs and text-block edges. Previous owner's inscription to front paste-down. Binding remains firm. Boards have mild edge-wear with slight rubbing to surfaces. Mild crushing to spine ends. Gilt lettering is bright and clear. Unclipped jacket has light edge-wear with tears and creasing.

Paperback

First published June 1, 1968

Reviews

[Steven Peterson · Rating 5 out of 5]

A colleague of mine at then the State University of New York at Buffalo gave me a copy of this book as I was working on my dissertation (in political science). I was at the start of my data analysis from surveys filled out by students. One standard approach was multiple regression, where one explores the effects of selected variables on the dependent variable, while controlling fr other variables that might be at play.

This book helped me make sense of one anomaly. The simple correlation coefficients showed no relationship between one key independent variable (psychophysiological arousal) and the dependent variable (political participation). However, when I ran multiple regression, all of a sudden the relationship between these two variables became statistically significant (large enough that it would be rare if chance alone could explain the relationship). I had, in fact, experienced one issue addressed by Rosenberg in this book--a "suppressor effect," in which one or more variables suppresses a genuine relationship between two variables. The book helped me to understand the dynamics of this quirky statistical matter.

So, count me as a satisfied consumer of Rosenberg's book.