Way back in the late '60's when I was a graduate student I was trying to master a now-antiquated technique called scalogram analysis. The procedure had been used by sociologists as quantitatively sophisticated as the late Paul Lazarsfeld in producing widely read research such as that included in the classic multi-volume work The American Soldier (1950).
Scalogram analysis is really a pretty simple technique for creating ordinal level attitudinal variables. It comes with a variety of measures, called coefficients of reproducibility, normed to range between 0 and 1, which purport to gauge the reliability of the variables created, such as tolerance of ethnic diversity. This technique always seemed like an iffy and unduly contrived way of measuring attitudes and creating variables to be used in statistical analyses. Still, it had the saving grace of reasonable looking coefficient(s) of reproducibility. When challenged, analysts could always look to a coefficient of reproducibility and say something like "Yeah, but it's pretty reliable. We got a coefficient of reproducibility value of .80 for a statistic that maxes out at 1.0"
In 1954, however, the psychologist Leon Festinger published an article in which he demonstrated that, given the way coefficients of reproducibility were calculated, values as high as .85 could easily be achieved just due to chance. One response to this challenge to the utility of coefficients of reproducibility, and, by implication, scalogram analysis itself, was publication of two articles in a 1959 edition the journal Psychometrica, one by Philip Sagi and the other by Leo Goodman, that presented tests of statical significance for some measures of reproducibility. I found the articles unreadable, and hoped to find a source that would provide clarification.
By chance, I stumbled across the first (1967) edition of Nunnally's Psychometric Theory. Nunnally's book did not address the issue of tests of significance for coefficients of reproducibility, but I was stunned by the clarity of Nunnally's prose. All of a sudden, measurement, a topic I had found unfathomable and deadeningly boring came to life. Nunnally was a gifted writer who made complex technical material accessible even to those of us who had little or no talent for that sort of thing. Reading it made me feel empowered.
Publishing textbooks is a money-making racket, so when Nunnally died, his publisher recruited a co-author, Bernstein, to periodically update Nunnally's book. Bernstein may be an unusually accomplished tests and measurement guy, but his prose rendering of technical material is not nearly as lucid as Nunnally's. I doubt that Bernstein tries to make measurement more difficult than it has to be, but he definitely lacks Nunnally's gift for making technical material understandable.
Psychometric Theory has gotten thicker, covering more material over the years. Nevertheless, if I were starting out or starting over, I'd try to find a copy of Psychometric Theory published before Nunnally's death in 1982, after which he posthumously acquired a co-author. Nunnally's first edition still provides a solid treatment of classical testing theory. Who knows? You may find all you need for your specific purposes in an earlier edition, saving yourself a lot of grief and, with any luck, a lot of money. Textbook publishing really is a racket.