Citation

Schoenberg R. Recent. Dev. Alcohol. 1985; 3: 289-300.

Copyright

(Copyright © 1985, Plenum Press)

DOI

unavailable

PMID

3975455

Abstract

The crucial problem in alcohol research is the large amount of inherently uncontrolled variation in the dependent variables. Because randomization with respect to key independent variables is frequently impractical or impossible, quasi-experimental and survey methods are being used to control residual variation statistically. For various reasons, however, the methods have often not been properly applied in alcohol research, endangering the replicability of true findings and prolonging the life of false claims. Ill-advised practices found in alcohol research that are discussed here are the use of partial correlations and standardized regression coefficients, of categorization or truncation of independent variables, and of available or convenient samples. It is recommended that researchers use unstandardized regression models, avoid categorizing independent variables, and scale them either to agreed on units of measure or use natural units (e.g., ounces or deciliters of alcohol consumed per occasion), and employ randomization or random sampling.

Language: en