NCJ Number
119491
Journal
Contemporary Drug Problems Volume: 15 Issue: 4 Dated: (Winter 1988) Pages: 565-606
Date Published
1988
Length
42 pages
Annotation
This essay points out that when alcohol and drug-related data are used to test hypotheses about causal relationships among phenomena varying in time, researchers must carefully control for the presence of trends in the data. Failure to do so may cause relationships between variables to be exaggerated, deflated or even inverted.
Abstract
Those carrying out sociological and epidemiological research often explain changes in one phenomenon in terms of changes in another phenomenon on the basis of comparison of trends. These explanations can be misleading, for two series may coincide even though there is no causal relationship between the two phenomena. On the other hand, two trends may be uncorrelated or divergent even though there is a positive and significant causal relationship between the two phenomena. Such anomalies occur frequently and it is important to distinguish apparent from sound evidence. The article asserts that a simple operation called filtering can solve some of these problems and outlines a statistical strategy for its use. 29 references.