Homework 3
This assignment pertains to material covered in Lectures 5-6. Please bring hard copy to class. 1 point for each item, 13 points total.
Question 1: Intimate Partner Violence.
A criminologist is interested in intimate partner violence (IPV). In particular, she wants to test whether victims of child abuse are more likely to experience intimate partner violence when they grow up. She collected a convenient sample among college students in China. The sample size was 2,000. The intimate partner violence was measured by a scale which summarizes the responses from whether the respondent had been the victim or perpetrator of the following types of abuse: (1) threats of violence, pushing or shoving, throwing objects that could injure a partner; (2) kicking, slapping or hitting; (3) sexual abuse (insisting on or making a partner have when he or she did not want to, or having a partner insist r make the respondent have sex when he or she did not want to); and (4) physical abuse that led to an injury, such as sprains, bruises and cuts. The child abuse is a dichotomous variable, with respondents coded as “1” if they responded affirmatively to two survey items about behaviors perpetrated by a parent or caregiver before the 6th grade, including: (1) being slapped, kicked or hit; and (2) forced to have sex. She also included a few control variables such as age, gender, parental educational attainment, parental income, and family structure.
IPV Perpetration
IPV Victimization
Parameter
Category
Beta(95% CI)
Beta(95% CI)
Child abuse
1.90(1.33 – 2.99)***
1.66(1.11 – 2.48)*
Parental income
Less than 2k
0.88 (0.73 – 0.96)+
0.67 (-0.39 – 0.95)
2k-5k
—
—
5k+
0.72(-0.42 – 0.40)
0.86 (-0.52 – 0.99)
Female
1.56 (1.21 – 2.00)***
1.53 (1.19 – 1.96)***
Age
-0.08 (-0.06 – 0.10)
-0.07 (-0.05 – 0.04)
Family structure
Single mother
-0.73 (-0.41 – 0.85)
-0.80 (-0.51 – 1.02)
Single father
2.83 (1.13 – 3.10)*
1.29 (-0.57 – 1.93)
Two non-bio parents
1.24 (-0.77 – 1.99)
1.23 (-0.82 – 1.86)
Two bio-parents
—
—
Notes: n=2,000. + p<.10; * p<.05 ; ** p<.01 ; *** p<.001. The intercepts were not shown.
a) Interpret the effect of Child abuse for IPV perpetration with 95% of confidence interval.
b) For a 15 years old girl living with low income father, what is the predicted value of IPV
victimization if she was not abused?
c) For a 15 years old boy living in a two bio-parents family with moderate income, what is
the predicted value of IPV perpetration if he was abused?
d) The criminologist submitted a paper based on the presented results to a journal, but one
of the reviewers pointed out the results may be invalid. Why is that? Hint: think about the way that data were collected.
Question 2: Gene and Environmental Interaction.
The genotype of an individual at a given position in the genome (called a locus) can be coded as a categorical variable such as AA, Aa, or aa, or indictor, e.g., AA=1, Other=0. Caspi A, et al. (2002) reported that people who carry a “2R” type in the
monoamine oxidase A (MAOA) gene promoter
have higher aggressive and antisocial behavior in adults if they were abused as children. In
contrast, individuals were abused as children but have another type in MAOA are less likely to
develop symptoms of antisocial behavior.
A researcher used a longitudinal study to replicate Caspi’s classical work. A group of high school
students were randomly selected, and repeatedly interviewed 4 times in 12 years.
Aggressive behavior is measured by a sum of the answers (5-point Likert scales) of the following
four questions.
In the past 12 months, how often did you
a. take part in a fight where a group of your friends was against another group? b. hurt someone badly enough to need bandages or care from a doctor or nurse? c. pulled a knife or gun on someone?
d. shot or stabbed someone?
MAOA gene is recoded into an indicator, 1=”2R”, and 0=”Other genotypes”. Mistreatment is measured by whether the respondent was sexually abused by his/her parents or other adult care- givers before 6th grade. Besides the MAOA gene and mistreatment, demographic, and other factors which are associated with aggressive behaviors are also included in the study.
In order to test the interaction effect, two models are estimated and organized in the following
table.
Model 1
Model 2
Parameter
Category
Beta(stderr)
Beta(stderr)
Intercept
1.361(0.077)***
1.362(0.077)***
Age
-0.143(0.012)***
-0.143(0.012)***
Race
Black
0.049(0.011)***
0.049(0.011)***
Hispanic
0.082(0.010)***
0.082(0.010)***
Asian
-0.002(0.016)
-0.002(0.016)
White
----
----
2 Bio parents
-0.030(0.008)***
-0.030(0.008)***
Female
-0.218(0.008)***
-0.217(0.008)***
MAOA
2R
0.053(0.029)+
0.033(0.029)
Other
----
----
Mistreated
Yes
0.171(0.019)***
0.162(0.019)***
No
----
----
2R× Mistreated
0.546(0.150)***
Social Support
-0.091(0.007)***
-0.091(0.007)***
Wave
1
0.100(0.031)**
0.100(0.031)**
2
-0.054(0.030)+
-0.055(0.030)+
3
0.179(0.018)***
0.179(0.018)***
4
----
----
AIC
130999.3
130988.0
BIC
131074.9
131063.6
N
49703
49703
Note: +: p<.10, *: p<.05, **: p<.01, ***: p<.0001.
Answer the following questions based on the table above.
a) According to Model 1, what is the estimated regression equation for White boys with 2R genotype and Mistreated at wave 1?
b) Interpret the effect of MAOA using Model 1.
c) Interpret the effect of Mistreated using Model 1.
d) According to Model 2, what is the estimated regression equation for the same group in 1)?
e) Interpret the interaction effect of MAOA and Mistreated using Model 2.
f) Is there any evidence that supports Caspi’s claim? Justify your answer using the
information provided.
g) Plot the interaction effect using Model 2. (Hint: you can do it in Excel!)
h) Comparing Model 1 and Model 2, which one of models does fit the data better? Why?
i) According to the information about how this replication was conducted, are there any
potential problems about this replication? For example, any violation of linear regression assumptions?