## ABSTRACT

The development of punched-card machines in the first third of the century made modem survey research economically possible. In this chapter, the authors provide the difference between contingency tables and partial correlation. In order to assess the relative merits of tabular analysis and linear statistical analysis the authors consider each set of procedures, concentrating on the logic of the procedures rather than on their application to delinquency research. They compare tabular analysis mainly with multiple regression, the best-known form of linear statistical analysis and the most useful in survey research. In tabular analysis, then, as long as the number of independent variables included in a table is smaller than the number known to be related to the dependent variable, the meaning of any relation is inherently ambiguous. Linear statistical procedures, programmed for large computers, are the aerial photographs of survey analysis.