Premium
IN SEARCH OF A DEPENDENT VARIABLE: COMMENT ON AVAKAME, 1998 *
Author(s) -
SCHWARTZ JENNIFER,
ACKERMAN JEFF
Publication year - 2001
Publication title -
criminology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 3.467
H-Index - 139
eISSN - 1745-9125
pISSN - 0011-1384
DOI - 10.1111/j.1745-9125.2001.tb00946.x
Subject(s) - homicide , spell , psychology , multilevel model , reading (process) , aggregate data , variable (mathematics) , aggregate (composite) , computer science , econometrics , poison control , mathematics , statistics , human factors and ergonomics , sociology , linguistics , medicine , machine learning , medical emergency , mathematical analysis , philosophy , materials science , anthropology , composite material
In a previous Criminology article, Avakame (1998) applies hierarchical linear modeling (HLM) techniques to Supplementary Homicide Reports (SHR) to disentangle individual‐ and aggregate‐level factors associated with offending. A close reading of his analysis reveals serious flaws in the dependent variable, which renders the results meaningless. Although it is ambiguous whether Avakame intended to model homicide “risk” or “frequency,” either is problematic. “Homicide frequency” has no logical connection to the individual‐level predictors; “homicide risk” is constant in SHR data, which makes the analysis impossible. In detailing these problems, we spell out the logical data requirements and offer sound empirical examples for an HLM analysis.