CHAPTER 3 (part 2)

THE WINNIPEG HEALTH AND DRINKING SURVEY - WAVE 1

Data Analysis

Data in this phase of the research were analyzed with the Statistical Package for the Social Sciences, version X (SPSSx) and the Statistical Analysis Systems (SAS). The following represents the general format of data analyses for this phase of the research:

1. Frequency analyses (using percentages and means) were conducted to describe the frequency and severity of perpetrated partner abuse as well as the demographic variables that define the sample.  At this stage of the analysis, it was also possible to test for scale outliers, skewness, normality, linearity and homoscedasticity and to conduct appropriate transformations when necessary.  Although the CTS (Straus, 1979) was found to be skewed, the criteria set by Cleary and Angel (1984) suggested the use of linear multiple regression was an appropriate statistical approach to analyze these data.  Finally, examination of the distributions of the alcohol dependence measures established their high/low cutpoints and facilitated the construction of the Alcohol Dependence Index.

2.Pearson Correlations were computed to determine the bivariate relationships among continuous or dummy coded variables.  As noted previously, correlational analyses determined the measures needed to be included in the Neuroticism Index.  In addition, zero order relationships between the dependent measure, partner abuse and the independent measures, socio-demographic, personality, alcohol measures were determined.

3. A number of standard regression models testing both main and interaction effects were performed on continuous and dummy coded variables.  These involved examining both the individual and combined effects of the independent variables against the dependent variable (i.e., six items taken from the CTS).
     Through this procedures, it was possible to ascertain which variables were most salient in their explanation of partner abuse.
 
 

Major Findings

The following represents a summary of the results of theanalyses conducted on the male and female data from Wave 1 of this research.
 

Male Data

1. The prevalence of male perpetrated abuse was 26.3 percent with the most common abuse tactic being "pushing, grabbing and shoving" (Barnes, Sommer, Murray & Patton, 1994) (See Table 2).
 
 

Table 2. Male perpetrated violence (Wave 1 data) (Barnes et al., 1994).

Type of violence		Number of occurrences	%
Minor violence acts
	Threw or smashed something (not at partner)	79	15.8
	Threatened to throw something (not at partner)	42	7.3
	Threw something at partner	30	4.6
	Pushed, grabbed or shoved	85	17.2
Severe violence acts
	Hit partner	42	7.3
	Hit partner with something hard	14	9
Violence Indexes
	Minor Violence	128	25.9
	Severe Violence	43	7.6
	Overall Violence	125	26.3

There were 10 missing cases. Overall violence scale statistics:

     Mean = 6.63, S.D. = 1.48 and range = 6-20

2. Partner abuse by male respondents was significantly predicted by being nonwhite, unemployed and alcohol dependent, and by a low score on Eysenck's Lie Scale and a high score on the Neuroticism Index.  In addition, an interaction effect was found for high alcohol consumption and high scores on the Neuroticism Index. This interaction effect was found to be the strongest predictor of male perpetrated partner abuse.  Table 3 illustrates the results of a standard regression model testing both main and interaction effects.
 
 

Table 3. Standard multiple regression predicting male perpetrated partner abuse (Wave 1 data) (Sommer et al., 1991).

* p<.05, ** p<.01, *** p<.001, F(19,311) = 5.92, p<.001,
adj. R2 = .22
 
 

Next: Chapter 3 Part 3