Blackwell Reference Online

Extract

Because of the difficulty in sociological research of conducting experiments that will yield valid and meaningful results, sociologists have looked for quasi-experimental designs in which statistical measures can be interrelated in a logical sequence to suggest causality. Path analysis is the most widely used of such approaches. Path analysis combines both a theoretical (or logical) analysis with a statistical analysis, inasmuch as a logical argument must be introduced to suggest appropriate causal sequences. While some of these logical connections are obvious (gender obviously comes before marriage, for example), others are a matter of theoretical argument as part of a larger project (does image of God precede or follow religious affiliation or are there ongoing interaction effects between the two?). Through multiple and partial regression analysis, and statistical controls that attempt to match test and control populations as closely as possible, a researcher attempts to demonstrate both quantitatively and logically a pattern of causality that would be similar to what might result if an actual experimental design were able to have been put into place by providing estimates of the total direct and indirect effects of one variable on another. In most simple causal modeling, path coefficients are beta weights , which represent a measurement of changes in a dependent variable in ... log in or subscribe to read full text