B4 - Applied Statistical Methods
Fall 2017
HWA Voting in Sweden
Johan Lyhagen/ Lars Forsberg
December 4, 2017
1 Introduction
The purpose of this home work assignment is to get practical experience in
modeling data using logit and probit models, both for binary and ordered data.
This includes estimation using the software package R, to intpret and analyze
the results.
HWA Voting作业代写、代写R语言作业、代做Statistical Methods作业、代做R课程实验作业
R is installed in computer room H319 and it is also possible to access through
remote desktop to ts03.dis.uu.se and log on as if you actually were in H319.
2 Written report
A written report should be handed in no later than the date indicated on the
schedule. The written report should be written in such a way that there is no
problem to understand what have been done and why.
3 Oral presentation
You shall present your work for the other students in ?seminar-format. Due
to the number of students taking this course, we need to split up in two seminar
groups, for simplicity, we will denote these groups:
1. Seminar group 1, and
2. Seminar group 2.
The dates and times for the two seminar groups are the schedule.
1
4 Group formation
Form working groups of two maximum three students. One of the students
send an email to
Lars:F orsberg@statistik:uu:se:
Please state in that email if you have any preference concerning which seminar
group you want to belong to. To the extent possible, these preferences will be
taken into account when allocating working groups to seminar groups.
You will then get a email-reply saying what
1. working group number you have,
2. what seminar group you belong to, and
3. what party you should analyze.
5 Data
The data used are from the European Social Survey which is a biennial multicountry
survey. In this home work assignment we are going to use data from
Round 3 and for Sweden, collected in 2006. In Table (1) the questions are
classi?ed into topics. A total of 487 variables on 1927 respondents are collected
but we are going to use a rather small subset. The subset of questions are
shown in Table (2): For a description of the formulation of each question see
the document ESS3Source_Mainquestionaire.
Question B12 is regarding which party the respondent voted for in last election.
The parties are as follows: the
1. Centre party (1)
2. The Liberal party (2),
3. The Christian Democrats (3),
4. The Green party (4),
5. The Conservative party (5),
6. The Social Democrats (6),
7. The Left party (7).
The codes 8-10 is for other small parties.
(As mentioned above, once you have formed the groups, you will be assigned
a party to analyze, in this presentation, we proceed as if we were assigned the
conservatives.)
2
Category Topic
A Media; Social trust
B Politics
C Subjective well-being; Religion
D Timing of life; The life course
E Personal and social well-being; Satisfaction (life, work)
F Socio-demographic pro?le
G Human values
H Test questions
I Interviewer self-completion questions
Table 1: Classi?cation of the questions of the European Social Survey, Round
3.
Topic Questions
A A1, A5
B B1, B11, B12
C C1, C5, C15, C28, C33, C35
D D4, D6, D9
E E4, E5, E7
F F1, F2, F3, F5, F27, F30, F32
Table 2: Variables availble in the ?le QDATA.wf1
6 Material on studentportalen
Material on studentportalen
3
7 Assignment
Below, the assignment is stated for the Conservative party, but as mentioned
above, you will be assigned what party to analyse.
1. Display and interpret descriptive statistics for how the voters voted in the
last general election in Sweden (variable B12).
2. Specify and estimate logit model for voters voting on your assigned party.
Dependent variable is "voting for the XXX party" (binary variable: 0,1).
The model should include a Likert scale explanatory variable. Justify/
motivate the model/ choice of variables. You should motivate why a specitic
variable is in the model and what sign you expect on the coe￠ cient.
Estimate and interpret the model.
3. Interpret the parameters of the model in a way you ?nd appropriate. Also,
for a typical individual of the sample (pick one), calculate the (model
implied) probability that this individual would vote for the conservative
party. Change the value of the explanatory variables and analyze how
the probability change. (What values of the other explanatory variables
would it make sense to use? Discuss and motivate)
4. Given the logit model in task 2: Test, using a likelihood ratio test, if
the (constant) scaling implied by the Likert scale, i.e. 1,2,3 etc is valid.
That is, is the restriction that the distances between the 1 and 2 are the
same as the distance between 2 and 3 etc valid? (You need to estimate
two models, one restricted and one unrestricted). What is the implication
of the test result in terms of the model and how to use the Likert variable
in this situation? Can this result be generalized, if so, how, if not, what
would be the recommended practice in this situation?http://www.daixie0.com/contents/15/2364.html
5. Make a two by two table of observed responses versus predicted. Use a 0.5
cut of value and then a cut of value corresponding to the unconditional
probability to vote for you assigned (XXX) party. Interpret and discuss.
Choose a goodness-of-?t measure to evaluate the model.
6. Using the probit instead of the logit, estimate the model in 2 and compare
the implied probabilities, how do they di§er? Compare the two by two
table of observed responses versus predicted for the two models. Which of
the Logit and Probit would you recommend to use for this data? Discuss
and motivate.
因为专业，所以值得信赖。如有需要，请加QQ：99515681 或邮箱：[email protected]
微信：codinghelp