1
Framing your research question,
constructing your hypotheses, and using an I.V. as a test factor.
4:1: How literature reviews produce propositions
4:2: Transforming propositions into hypotheses, the formula for hypothesis
construction, with nulls and rationales
4.3: Recoding/ collapsing / dichotomizing variables
4:4: Adding test factors/ control variables
4:5: Writing the research question & hypothesis section of the paper
Vocabulary/Concepts:
Hypothesis
Hypothesis rationale
Collapsing/recoding
Independent variable
Null hypothesis
Dichotomous variables
Dependent variable
Control variable/test factor
2
4:1) How literature reviews produce propositions:
Literature review: A review of relevant literature indicates that support for capital punishment varies by
numerous sociodemographic characteristics. Among these characteristics, gender and political
identification emerge as variables that have been found to be consistently associated with differences in
support for the death penalty. More specifically, the literature indicates that men and political conservatives
are more likely to support the death penalty than women and liberals (Jones 2000; Stack 2000; Ellsworth
and Gross 1994).
Proposition: Support for capital punishment varies by gender and political orientation.
Preliminary research question: In keeping with prior research findings, does support for capital
punishment vary by gender and political orientation in the 1996 GSS sample?
Process:
1. We have an interest in attitudes toward capital punishment (=D.V. selection).
2. We reviewed the literature to acquaint ourselves with past research on DPS.
3. We sought out a data set that provided DPS public opinion data (= GSS selection). DPS is
CAPPUN variable in GSS96TAB.SAV data set.
4. We isolated key independent variables that were found to be important in prior research (= array
of possible I.V.s on chart)
5. The literature indicates gender is an important I.V. and SEX is available in GSS96TAB.SAV
(=selection of I.V.).
6. The literature indicates political orientation is an important I.V. and POLVIEWS is available in
GSS96TAB.SAV (=selection of I.V.).
4:2: Transforming propositions into hypotheses
4:2a) Construct Hypothesis #1:
Variable
Concept
D.V.
DPS
I.V.
Gender
GSS variable
CAPPUN
SEX
Values/Attributes
Favor=1 Oppose =2
Male=1 Female=2
Hypothesis formula:
(I.V.a1) is more/less likely to (D.V.a1) than (I.V. a2).
Hypothesis:
Men are more likely to support the death penalty than women.
Null hypothesis:
There is no difference between men and women in their support of the death penalty.
Rationale: (Insert prior research that leads to this expectation. Make a concise summary paragraph citing
research in which men were found to express greater DPS than women. You may also include explanations
posed for gender differences in DPS).
Both the first I.V. of gender and the D.V. of DPS are dichotomous. That is, they have only 2 values or
attributes: Gender: male / female; DPS: favor / oppose
If we examine how one is related to the other, we have a 2 x 2 table.
Gender
male
female
favor
DPS
oppose
3
4:3) Construct Hypothesis #2: Recoding / collapsing / dichotomizing variables
A dichotomous I.V. & D.V. also make it very easy to write a hypothesis using the formula because you
have the right number of attributes to fit into the formula. However, POLVIEWS in the GSS96TAB has 3
attributes.
Variable
D.V.
I.V.
Concept
DPS
Political orientation
GSS variable
CAPPUN
POLVIEWS
Values
Favor =1
Oppose =2
Liberal=1 Moderate=2
Conservative=3
So, now what?
Now, you have to decide how to collapse the three attributes into two attributes.
Options:
a) Liberals & Moderates v. Conservative
b) Liberals & Conservatives v. Moderates
c) Liberals v.
Moderates & Conservatives
(note: excluding one attribute is an option, but we will not be doing that at this point.)
Since more of the literature centers attention on conservatives versus all others, we will select the first
option in which Conservatives are isolated from the other two groups. However, since Conservatives are
our focus, it is more convenient to assign a value of 1 to Conservatives and a value of 2 to what we will
now call Non-conservatives.
POLVIEW2 : Conservatives=1 Non-conservatives=2
note: I’ve given the recoded variables a new label-ending with 2 to denote a recoded variable.
In the lab, you’ll learn how to do this recoding process.
Construct Hypothesis #2:
GSS data set selected:
Variable
D.V.
I.V.
Hypothesis:
Null hypothesis:
Rationale:
Concept GSS variable
Values
4
4:4 Construct Hypothesis #3: Adding a control variable or test factor.
With two I.V.s, you have created two separate hypotheses. However, you can use one I.V. as a test factor
or control variable. This is how we do this:
You posed that men are more likely to support the death penalty than women and that
conservatives are more likely to support the death penalty that other political orientations. It is logical to
expect that the two I.V.s might work together to create an even stronger association. In this case it is
reasonable to expect that if men and conservatives are more likely to support the death penalty, then
conservative men are more likely to support the death penalty than conservative women.
Variable
Concept GSS variable
Values
D.V.
I.V.
C.V.
Hyp. W/ control variable (C.V.): Men are more likely to favor the death penalty than women,
and this is more likely to hold true for conservatives than for those of other political orientations.
Null hypothesis:
Rationale:
Only respondents who said they are conservative are used in the analysis.
Conservative
Gender
male
female
favor
DPS
oppose
Only non-conservative respondents are used in the analysis.
Non-conservatives
Gender
male
favor
DPS
oppose
Note: Your first I.V. and D.V. form the basic matrix.
female
5
3:6) Writing the research question & hypothesis section of the paper
Research Question & Hypotheses
[Section Heading]
[Subheading #1]
Introduction (Brief literature review overview statement )
Sample: A review of relevant literature indicates that support for capital punishment varies by numerous
sociodemographic characteristics. Among these characteristics, gender and political identification emerge
as variables that have been found to be consistently associated with differences in support for the death
penalty. More specifically, the literature indicates that men and political conservatives are more likely to
support the death penalty than women and liberals (Jones 2000; Stack 2000; Ellsworth and Gross 1994).
[Subheading #2]
Research question:
In keeping with prior research findings, does support for capital punishment vary by gender and political
orientation in the 1996 GSS sample?
[Subheading #3]
Hypotheses
[Sub-subheadings]
Hypothesis # 1:
Null hypothesis #1
Rationale (with literature refs)
Hypothesis # 2:
Null hypothesis #2
Rationale (with literature refs)
Hypothesis #3:
Null hypothesis #3
Rationale (with literature refs)
[Subheading #s]
Summary