president. Yet Democrats have been the majority party in the House of Representatives throughout this period, and the majority in the Senate for all but six years of the period. An analysis of Presidents' statements accurately reflects the inclinations of the executive branch through this period, but is far from accurate in reflecting the inclinations of the legislative branch.
Likewise, if we wished to analyse the inclinations and biases of Hollywood's "creative community," we would encounter one sampling problem if we selected for content analysis the top grossing films of the 1980s, and a comparable but somewhat different sampling problem if we selected the Oscar nominees for the same period. They are different films, chosen by different people for different reasons. Neither subgroup is representative of all films released in the decade, much less of the vastly larger number of screenplays or treatments produced during the period.
Within a given document, we must also be concerned about our methods of sampling words, phrases, or on camera appearances, much less broader and more nebulous concepts such as "themes." Discussions of movie violence last year sometimes noted the "body count," reporting for example that "Die Hard II" had something like 350 on screen killings. In fact, however, the majority of those killings were in the crash of a single jumbo jet practically speaking, a single violent episode, and not really comparable to seeing the same number of people gunned down one or two at a time. By the same logic, one could say that the original "Superman" movie hardly Hollywood's most violent by our intuitive standards was the all time slaughter champion, because it showed the destruction of the planet Krypton and its entire human like population of many billions.
The examples given above are somewhat extreme, but they demonstrate the problems of "sample selection" in content analysis....