You are on page 1of 6

1

Ordered Logistic Regression of Effects of Level of Religious Knowledge on Perceptions of


Dependence of Family Honor on Female Sexuality
SOC 6233 - Advanced Sociological Statistics: Homework 4
Rafia Mallick
October 1, 2017

1. Question
This study assesses the relationship between Level of Religious Knowledge (LRK) and the
perceived association between female sexuality and family honor (FSFH). Commented [C1]: I think I like this better.

2. Data
The data for my analysis comes from the OU Honor Crime Study (OU HCS) 2017. The
sample comprises of all US-born and international students who enrolled in undergraduate and Commented [C2]: Can you just say all?
graduate level programs in the Summer and the Fall semesters at the University of Oklahoma
(OU), Norman Campus in 2017. The respondents were recruited using the OU mass email sent to
all enrolled students. A sample of 1,786 (8.9%) respondents participated in the study from the total
population.1 While the response rate is quite low, I argue that the information analyzed here Commented [C3]: This info is more suited to a note I think
the language should also be tightened up
remains relevant for X, Y, and Z reasons. The mass mail recruitment technique tends to have
relatively low response rate compared to traditional methods (Nulty, June 2008). Various studies Commented [C4]: For you this is an absolute must You will
bolster the claims and address an immediate question at the
(e.g. (Schuldt & Totten, 1994; Tse, et al., 1995; Weible & Wallace, 1998) have found that mass frontend.
mail has lower response rates than the postal mail. Response rates for online surveys vary from
Potential reasons - exploratory, hard to collect any info on these
about 6 percent (Tse, et al., 1995) to almost 75 percent (Kiesler & Sproull, 1986). While low topics, no extent data available, etc.
response rates are an issue, because many student studying at OU block the mass mail address due
to too many emails being received through it on daily basis, also many students ignore many
research emails being received as they do not usually fit in the criteria to participate in them.
3. Dependent Variable
In order to measure FSFH, I implemented the Understanding of Honor Scale taken from
the work of Isik (2008). This scale is concerned with capturing the respondents attitudes toward
concerning whether dependence of family honor is dependent on female sexuality. The scale uses Commented [C5]: Maybe on the sexuality and sexual
practices of female family members
15 items measured on likert scale of Strongly Disagree to Strongly Agree. As an example, one
of the first question in this scale asked: Women's virginity should be kept under control for the This is reading cleaner, but I would continue to play with the
protection of honor (See appendix XXX for a full list of questions in this scale). Six of the items language.

were later reverse coded (Item number 4, 6, 7, 13, 14 and 15) as they were reverse statements
asking if the respondents were against the attitude that the female sexuality is a determinant of
family honor, while all others are direct agreement statement.2 (See Appendix for the scale) With
these fifteen items, I create a scale by summing them all together into a single measure of FSFH.

1
Overall, the response rate was around 8.9 percent of the total population. The mass mail recruitment technique
tends to have relatively low response rate compared to traditional methods (Nulty, June 2008). Various studies
(e.g. (Schuldt & Totten, 1994; Tse, et al., 1995; Weible & Wallace, 1998) have found that mass mail has lower
response rates than the postal mail. Response rates for online surveys vary from about 6 percent (Tse, et al., 1995)
to almost 75 percent (Kiesler & Sproull, 1986). WhileThese low response rates are an issuemay be due to , because
many the large number of OU mass email a received by student studying at OU block the mass mail address due to
too many emails being received through it on daily basis, also many students ignore many research emails being
received as they do not usually fit in the criteria to participate in them.
2
One of the items asked; Male relatives should not interfere with a woman's honor.
2

Overall, these fifteen items form a very reliable ( = .87) scale, suggesting that these statements
are an internally consistent measure of FSFH. Commented [C6]: Better here But not a finished product yet.

The final variable measure is d the combined attitude that the female sexuality is a
determinant of family honor and was on a continuous scale varying between a possible -30 and
+30. For the purposes of this homework, I recode this This variable was again recodedinto to a
likert five category ordered scale. Here values for -30 -16 represent strong disagreement (code Commented [C7]: This is not possible
0, values XX-XX represent general disagreement (coded 1), etc. where the values ranged from 1
to 5, indicating strongly disagree to strongly disagree to strongly agree that family honor is
dependent on female sexuality.3 About 93.53% of the respondents (N=1,300) were inclined Commented [C8]: Cant footnote this
towards disagreement with the dependence of family honor on female sexuality. Commented [C9]: This does not quite work for me be clear
and thorough in your description
4. Independent Variable:
In order to measure religious knowledge, I implemented the Centrality of Religiosity
Scale (CRS) taken from the work of Huber and Huber (2012). The scale uses 3 items4 and measure Commented [C10]: Cant footnote this info
on a likert scale of strongly disagree to strongly agree. Overall, these three items form a somewhat
reliable ( = .77) scale, suggesting that these statements are a consistent measure of level of
religious knowledge. The new variable formed varied from 0 to 12 on a continuous scale.
Sex of respondent was measured by a closed ended question, asking respondents How do
you identify yourself? The with the fixed categories included including male, female, transgender Commented [C11]: As a professor I hope this is not the exact
language you used because this is an issue.
male, transgender female and an open ended category other. It is recoded as a dummy variable
such that males are coded as 1 and all others as 0. Almost 40.79% of sample are males. Commented [C12]: This is theoretically not great You have
some choices here I think the best is all who identify as male
Respondents religion religious belonging was measured by a closed ended question, (either male or male trans) are coded 1 and all who identify as
female (female or female trans). The others you may want to omit.
asking respondents Do you now see yourself as belonging to any particular religion? A number
of religion categories5 were provided along with an open ended category other. It is recoded as You asked for information here that makes coding much more
a dummy variable such that Christians are coded as 1 and all others as 0. Almost 51.65% of sample difficult and due to the small number you cannot run any
meaningful statistics on these groups. A sensible coding is needed
are Christians. here.

Sexual orientation was measured by a closed ended question, asking respondents How do Commented [C13]: Seems low to what I know about this
population.
you define your sexual orientation/sexual identity? The categories included heterosexual or
straight, homosexual, gay or lesbian, bisexual and an open ended category other. It is recoded
as a dummy variable such that heterosexual is coded as 1 and all others as 0. Almost 78.6% of
sample are heterosexual. Commented [C14]: You are not providing enough information
as a rule here for the homework, chose a couple of controls and
leave the others for the paper That way you might have the space
to write up the results (which is the main goal of the HWs)
5. Analytical Strategy
The study uses ordered logistic regression to estimate the relationship between all
independent variables and FSFH. Model 1 in the table below shows the bivariate relationship Commented [C15]: WHY
between religious knowledge and FSFH. Demographic variables were added into Model 2,
showing a multivariate relationship between the level of religious knowledge, sex, sexual

3
-30 to-16 was coded as "Strongly Disagree, -15 to-1 as "Somewhat Disagree", 0 as "Neither Disagree nor Agree",
0 to15 as "Somewhat Agree" and 16 to 30 as "Strongly Agree".
4
The questions ask I am interested in learning more about religious topics, How often do you keep yourself
informed about religious questions through radio, television, internet, newspapers, or books? and How often do
you think about religious issues?
5
Agnostic, Atheist, Christian, Muslim, Jewish, Hindu and Buddhist.
3

orientation, religion and FSFH. Model 3, adds an interaction term of LRK with sex of respondent
to the multivariate model.
6. Results
[FIGURE 1 ABOUT HERE]
The results in Figure 1 show the probability of predicting the perceptions of dependence of
female sexuality on family honor (FSFH). In the first respondents who strongly disagree with
FSFH, we see the probability of predicting strongly disagreeing with FSFH is highest when the
respondent has no religious knowledge. The probability of predicting somewhat disagreeing with
FSFH is highest, at around 42%, when the respondent has high religious knowledge. The
probability of predicting neither disagree nor agree with FSFH is almost same across religious
knowledge. The probability of predicting somewhat agreeing with FSFH is highest, at around 16%,
when the respondent has high religious knowledge. The predicted probability of strongly agreeing
with FSFH remains constant across level of religious knowledge.
[TABLE 1 ABOUT HERE]

You did not complete this and you turned it in late. It might not be a great idea to wait until the
weekend to complete your HW. While the HW that Mitch and Martin assign lends itself to collaborative
work, the HW in my class do necessarily do so. You may want to start during the week and then when
you work in the office on the weekend you and your collaborators can read each other work. If I were to
guess, you are writing these last minute and this might explain some of the short coming I am noticing. I
am particularly worried that you have not given yourself the space to interpret your tables You might
reduce the number of controls so you do not have to spend the space describing these
4

Commented [C16]: I am not sure what is happening here the


titles of the cells needs to tell the reader something
Commented [C17]: No color please
5

Table 1: Ordered Logistic Regression of Effects of Level of Religious Knowledge on


Perceptions of Dependence of Family Honor on Female Sexuality Commented [C18]: This look(much) better, than your previous
table format Let have this be the standard

Model 1 Model 2 Model 3

Level of Religious Knowledge 1.16*** 1.10*** 1.16*** Commented [C19]: Use some sensible indentation on the
coefficients to tell your story
(-0.02) (-0.02) (-0.04)
Commented [C20]: This is also a concern. Standard errors can
Christian 2.75*** 2.76*** never be negative
(-0.37) (-0.37)
Male 2.52*** 4.55***
(-0.31) (-1.25)
Heterosexual 2.94*** 2.89***
(-0.56) (-0.55)
Male*Level of Religious Knowledge 0.91*
(-0.04)
Somewhat Disagree Intercept 4.14*** 20.32*** 27.63***
(-0.53) (-4.46) (-7.17)
Neither Disagree Nor Agree Intercept 34.79*** 204.50*** 276.80***
(-5.77) (-52.00) (-79.87)
Somewhat Agree Intercept 42.99*** 254.40*** 343.80***
(-7.45) (-66.05) (-100.80)
Strongly Agree Intercept 849.60*** 5172.10*** 6958.40***
(-439.10) (-2858.80) (-3957.20) Commented [C21]: These are a problem Something has gone
wrong here.
Formatted Table
Observations 1390 1390 1390
AIC 2344.30 2160.40 2156.60
BIC 2370.50 2202.30 2203.70

Exponentiated coefficients; Standard errors in parentheses


Source: OU Honor Crime Study
* p<0.05, ** p<0.01, *** p<0.001
6

Commented [C22]: Nice inclusion


Bibliography
Cihangir, S. (2013). Gender specific honor codes and cultural change. Group Processes & Intergroup
Relations, 319333.

Fisk, R. (2010, September 7). The crimewave that shames the world. Retrieved November 17, 2016, from
Independent: http://www.independent.co.uk/voices/commentators/fisk/robert-fisk-the-
crimewave-that-shames-the-world-2072201.html

Huber, S., & Huber, O. (2012). The Centrality of Religiosity Scale (CRS). Religions, 710724.

IIK, R. (2008). The Predictors of Understanding of Honor and Attitudes Toward Honor Related Violence:
Ambivalent Sexism and System Justification. Ankara, Turkey: MIDDLE EAST TECHNICAL
UNIVERSITY.

Kiesler, S., & Sproull, L. (1986). Response effects in the electronic survey. Public Opinion Quarterly, 402-
413.

Nulty, D. (June 2008). The adequacy of response rates to online and paper surveys: what can be done?
Assessment & Evaluation in Higher Education, 301314.

Schuldt, B., & Totten, J. (1994). Electronic mail vs. mail survey response rates. Marketing Research.
Marketing Research, 1-7.

Tse, A., Tse, K., Yin, C., Ting, C., Yi, K., Yee, K., & Hong, W. (1995). Comparing Two Methods of Sending
Out Questionnaires: E-Mail Versus Mail. Journal of the Market Research Society, 441-446.

Weible, R., & Wallace, J. (1998). Cyber research: The impact of the Internet on data collection.
Marketing Research, 19-31.

(2015). When Men Murder Women - An Analysis of 2013 Homicide Data. Washington, D.C.: Violence
Policy Center.