Insights into Regression Analysis
Introduction
REGRESSION ANALYSIS 2
2
The regression technique used is the logistic regression. This is a technique used when
the dependent variable is categorical. The study outcomes are not continuous and they do fall
into different categories that warrant this modeling approach (Lewis-Beck, 1993). In a logistic
regression, a logistic transformation of the odds or the logit serves as the dependent variable. The
dependent variable (odds) if taken and adding a regression equation for the independent variable
will yield a logistic regression (Allen, 1997).
The odds ratio is used to measure how the fitted probability changes between different
values of the explanatory variable. An odds ratio that is equal to one indicates a zero effect, an
odds ratio greater than one indicates an increase in odds while an odds ratio less than one
indicates a decrease in the odds (Draper, & Smith, 1981).
A logistic regression significance testing helps a researcher to identify the influence of
different factors in his model (Lewis-Beck, 1993). Thus, the Wald’s test is used to test the
contribution of individual regression co-efficient, while the deviance test is used to identify the
contribution of several model coefficients simultaneously.
Criminal Justice Research
The research has looks into different types of mistreatment categorically. The odds ratio
with its confidence interval has been indicated, as well as the regression co-efficient for each
contributing factor. The Wald value has also been indicated so as to ascertain the contributory
aspects of each variable. The p-value to indicate the significance of each variable has also been
indicated.
REGRESSION ANALYSIS 3
3
In analyzing emotional mistreatment, age, employment status, health status, previous
traumatic event, social support levels, and ADL assistance needs are our categorical variables
with a probability of success or failure (Williams, 1959). From the results, the odds of an a
person below seventy years being mistreated emotionally is approximately three time more than
for a person above seventy years. The regression coefficient also indicate that age is the most
influential factor when understanding aspects of emotional mistreatment followed by previous
traumatic event, need for ADL assistance, health status, employment status and social support
respectively. Nonetheless the health status is insignificant when interpreting the health status.
Social support level contributes more to the model individually.
Physical mistreatment can be associated with age, gender, race, income level, health
status, previous traumatic event and level of social support. Age and social support levels are the
only significant factors that can be used to define physical mistreatment as indicated by their
probability values of 0.05. The model also suggests that age is the most influential factor as
regards physical mistreatment. This is attributed to its Wald value of 8.5 and regression
coefficient value of 1.41.
Sexual mistreatment as a categorical variable is identified by gender, income levels,
health status, previous traumatic event, social support and need for ADL assistance. Of these
factors, previous traumatic event is the only significant variable explaining sexual mistreatment.
It regression coefficient of 2.64 indicates its high influence on the logistic model describing
sexual mistreatment.
The research utilizes logistic regression to assess different aspects of elderly mistreatment
based on different predictors ascribed to them. This is due to the categorical nature of the various
REGRESSION ANALYSIS 4
4
forms of mistreatment. Mistreatment is our outcome variable as its value is discrete, thus the
assumption of linearity is upheld. A logistic regression then addresses this defect by using a
logarithmic transformation on the type of mistreatment variable. The different levels of
mistreatment are expressed in terms of odds ratio for success. This ratio indicates the reaction
resulting from any form of mistreatment.
References
REGRESSION ANALYSIS 5
5
Allen, M. P. (1997). Understanding regression analysis. New York: Plenum Press.
Draper, N. R., & Smith, H. (1981). Applied regression analysis. New York: Wiley.
Lewis-Beck, M. S. (1993). Regression analysis. London: Sage Publications.
Williams, E. J. (1959). Regression analysis. New York: Wiley.
