CHAPTER 4
The main aim of this study was to determine if there was a statistical significant difference in how men and women perceive the primary aggressor, the more likely to sustain injuries, the appropriate punishment and who is likely to suffer the greatest trauma in domestic violence. Furthermore, the study also examined if there was gender based difference in individual variables such as, primary aggressor, injuries sustained, form of punishment and greatest trauma suffered. To achieve this, the analysis was divided into four parts, that is, (1) descriptive statistics; aimed at providing information on the collected data. (2) Reliability and factor analysis; was used to show inter-consistency among the variables given that they came from multiple Likert questions. (3) Correlational analysis; was used to show relationship between the variables and if correlation was statistically significant. (4) Inferential analysis; these was used to test the hypotheses earlier stated through various statistical tests.
Descriptive Statistics
Descriptive statistics refer to the science of describing the major attributes of a collection of day (Cope, 2008), or better referred to as quantitative description. They tend to summarize the sample under study rather than infer about the population (Shipway, 2004). Our study involved 200 participants (100 females and 100 females) aged between 18-50 years. All participants took part in the study and thus there were no missing values.
Various means and standard deviations were computed in SPSS and summarized in the table below. They were found to be of importance since through the means, the study was able to tell which group was higher given the predictor variables under investigation. For example we could tell that female had a higher scored highly (2.52) on suffering under trauma than men (2.39). Though not sufficient for making a conclusion, it provides a rough picture of what is expected in the inferential statistics once we realize the is a significant statistical difference (Cowman, 1998).
Reliability analysis
Reliability analysis is performed when a study involves multiple Likert questions that form a given scale. It involves measure of internal consistency (reliability) determined by computing the Cronbach’s alpha which represents its coefficient ( Novak, 2004). Reliability can be said to refer to the consistency and stability in the results of a test or scale. A scale would be said to be reliable if it outputs same results when administered repeatedly given that the attribute in check remain constant in the intervals between measurements. For example a weighing machine which gave 130 kg then a second time gave 140kg would be termed unreliable scale for measuring weights. In addition, an instrument must be confined to measure a single attribute and only in one dimension. For example, if a questionnaire set to measure anger measured stress at the same time, it would not be a reliable measure of anger. This study made use of multiple Likert scale ranging from strongly agree, agree, disagree and strongly disagree. As a result reliability analysis was carried out to determine the reliability of this scale. Further, factor analysis was used in determining if the instrument used (questionnaire) used a single dimension.
The alpha coefficient for the four items was 0.688 which tends to 0.7, thus the items under study were considered to have and acceptable level of internal consistence. Further, the study used factor analysis to determine the dimensionality of scale and how the variables were extracted. As per the Total Variance Explained table, the study found out that there was a large difference between the eigen value of the first factor and that of the second factor (2.198 vs. 0.8). The first factor also accounts for 54% of the total variance. Thus, the study concluded that the scale items were un-dimensional (hence measures only a single dimension).
Spearman rank correlation coefficient
Correlation simply measures the strength between two variables. These variables tend to be continuous (Noaks, 2009). Since this study made use of gender which is a dichotomous variable, of interest was not the linear strength of correlation but rather if the correlation was statistically significant.The spearman correlation coefficient, was named after Charles Spearman, refer to a non-parametric measure of dependence between variables. It tries to determine how well the relationship of two variables can be told through a monotonic function. It ranges between -1 to 1 with values on extreme ends indicating greater monotonic relationship between variables and values close to 0 indicating weak monotonic relationship. This study needed to taste if the correlation was significant where at least a part of the independent variable would statistically be explained by the dependent variable. For example, did gender to a given level affect the primary aggression?
Spearman correlation coefficient (Katz, 2006) was used to show the strength of the relationship between the variables and gender. For the primary hypothesis, index was correlated with gender and a correlation coefficient of -0.247 was determined. This indicated a week correlation. However, it was noted that the coefficient was statistically significant and thus appropriate to use. For the secondary hypothesis, the spearman correlation coefficient was still used where by, most of the variables had weak negative monotonic relationship (aggressor = -0.184, injury= 0.373, punish= 0.270, trauma = 0.137) however, it was seen that all of the correlation were statistically significant (p<0.01) apart from trauma (p>0.01). This gives us an idea that actually there might not be any difference between men and women with respect to suffering due to trauma as a result of gender based violence.
Hypothesis testing
Primary hypothesis that was being tested was that there will not be a significant difference in men and women perceptions as to the primary aggressor, sustenance of injuries, appropriate punishment and the greatest victim of trauma in domestic based violence. To achieve this, the major tests used were the independent sample to t-test and one way ANOVA to show if there was a statistical significant difference. One way ANOVA formed a larger part of testing the hypotheses given that it’s more flexible and has less prior conditions (assumptions).
Just to understand this test, the student t-test refers to a statistical hypothesis test whose test statistics tend to follow a student t-distribution under the null hypothesis. It is used to determine if two sets of data differ significantly from each other. This occurs especially in the independent sample t-test when two separate data sets are derived from same population or different populations but have the same distribution. The study chose to use the t-test given that the variables followed a normal distribution and thus men and women formed two independent identically distributed data sets (figure 1.0). On the other hand, one way ANOVA test, is a statistic test used to compare means of two or more samples using the Fisher’s distribution. Basically when dealing with two groups, it becomes an extension of the t-test. This study used one way ANOVA due to its flexibility where more than two variables are concerned. It follows the same assumptions as the t-test stated above and verified using figure 1.0 below.
Figure 1.0
The independent sample test showed that there was a significant difference (p<0.01) between men and women with respect to their perceptions of gender based violence in terms of aggressor, injuries sustained, punishment to perpetrators and suffering due to trauma. Due to the assumption of variance equality in the t-test and its complexity and being detailed, one way ANOVA was also used to test these hypothesis where (p<0.01) and thus the study concluded that indeed there was a significant difference between the male and females.
Secondary hypotheses
Hypothesis 1;
The first hypothesis stated that; there will not be a statistically significant difference in how men and women perceive who the primary aggressor in domestic violence cases is as the null hypothesis. In testing these hypothesis, the study used one way ANOVA where it was found out that there was a significant difference ( p<0.01) between men and women (table 1.9). Thus the study concluded that men and women perceived the primary aggressor in domestic violence differently.
Hypothesis 2;
The second hypothesis stated that; there will not be a statistically significant difference in how men and women perceive who is more likely to sustain injuries in domestic violence addressed by question 6 and question 7. ANOVA test was used and it was determined that there was a significant difference (p<0.01) (table 1.9) thus the study concluded that men and women perception of who was more likely to sustain injuries was significantly different.
Hypothesis 3;
The third hypothesis stated that; there will not be a statistically significant difference in the choice of the appropriate choice of punishment for domestic violence cases between men and women. The ANOVA test showed that there was a statistically significant (p<0.01) difference in the choice of punishment between men and women (table 1.9).
Hypothesis 4;
The last hypothesis stated that; there will not be a statistically significant difference in how men and women perceive who is likely to suffer the greatest trauma for domestic violence addressed by question 11 and question 12. ANOVA results showed that there was no statistically significant difference ( p>0.01),(table 1.9) thus we can conclude that men and women have the same perception of who suffers the greatest trauma in domestic violence cases.
Summary
The use of descriptive statistics was helpful in that it provided the quantitative description of the variables. The study was able to deduce that there were no missing values by looking at the frequency table for gender. From the means, it was possible to tell which group between the men and women performed better. The reliability and factor analysis came in handy, it was of paramount importance for the study to determine that amid using multiple Likert questions, the data was still reliable for further analysis. The factor analysis were also important in realizing the variability explained by each level. And this was the case since through the results, it was evident that the Chronbach’s coefficient was acceptable and the scale was un-dimensional.
The spearman coefficient was helpful in determining the strength of the variables under study. With coefficient ranging from -1 to 1 (Mandel, 2012), we were able to tell that most of the variables had a weak collinearity respect to gender. The independent sample t-test was good since we were dealing with two independent groups. It helped in easily deducing the difference in the two groups. However, to use t-test meant that each time a new hypothesis was to be tested, a new table would have to be created. So to ease this, the study adopted the one way ANOVA which basically works the same way like an independent sample t-test since we know from statistics that the square of t-test will generate an F-test (Denscombe, 2007).
The secondary hypothesis is tested using four parts, as per the first hypothesis; there is a significant difference on how men and women perceive who the primary aggressor in domestic violence is. Based on the mean, men have a higher score than women (2.73 vs 2.58) thus we can now further conclude that men tend to be more of primary aggressors than women.
The second hypothesis, there is not a significant difference in the way men and women perceive who is more likely to sustain more injuries in domestic violence, we see that there is a significant difference. In this case the men score higher than the women (2.65 vs 2.42) thus we can conclude that men tend to sustain more injuries in domestic violence than women.
The overall conclusion for this study would be there is a significant difference in how men and women perceive various aspects of gender based violence. Given the descriptive statistics though, we would be right to conclude that men tend to always be on a high Likert scale when attributes of gender based are of concern. The study showed that men are the top primary aggressors and hence tend to initiate most gender based violence. These concurs with previous studies that most gender based violence is directed towards women. However, men take up a high percentage when it comes to being victims of injuries as a result of gender based violence. This is a contradiction and maybe calls for closer and deeper analysis on the same. So as per this study research was left at determining if there was any statistical difference in how men and women perceive aspects of gender based violence but it would be prudent to further research on the same and determine exactly on what scale difference do the two groups differ.
References
Cope, N. (2008) ‘ ‘Interpretation for action?’: definitions and potential of crime analysis for policing’ Ch16 in Newburn, T. (2008) (ed) Handbook of Policing 2nd edition. Cullompton: Willan Publishing
Cowan. G (1998) Statistical Data Analysis Clarendon Press
Denscombe, M. (2007) The Good Research Guide for small-scale social research projects. Third edition. Maidenhead: Open University press
Katz. M (2006) Study Design and Statistical Analysis Cambridge University Press
Mandel. J (2012) The Statistical Analysis of Experimental Data Courier Dover Publications
Noaks, L. Wincup, E. (2009) Criminological research. Understanding Qualitative methods. London: Sage Publications ltd
Novak. C (2004) The Oxford Dictionary of Statistical Terms. John Wiley & Sons Ltd
Shipway, L. (2004). Domestic violence: A handbook for health professionals. New York: Psychology Press.
