2020 | Home assignment | retake exam
BA-BPOLO2010U: Quantitative Methods for Business and Social Science
General instructions
Content and format:
• The assignment is individual and should be maximum 10 CBS standard pages. Please follow CBS
guidelines regarding all content.
– This is an upper limit, fewer pages are fine.
– The R-code that you provide in the Appendix does not count towards your page limit.
• Do not include the text of the questions in your answer submission, but do include the exact
question number. Submit the final document with the subquestions following the order from the
exam sheet.
• Respond to all questions to the best of your knowledge and do it in your own words.
• Do not plagiarize and keep in mind that this is an individual assignment!
• In case you are using a particular idea/comment from somewhere, please add a reference. Use any
style, but be consistent (Chicago, APA, etc.).
• If you are asked to calculate something, describe how you calculated it and put the results in context.
Elaborate on your answers in order to show that you are able to explain statistical results and able
to make substantive interpretations.
• Remember to report relevant output in the exam document. You can just copy-paste it directly from
R or add the figures, you do not need to spend much time on formatting tables for this exam as long
as the results you present are readable and understandable.
• Everything must be handed in as one .pdf file (including appendix of R code appendix).
Notes on data and R code as appendix:
• All data needed is available on the course Canvas page.
• For answers that involve data analysis, you need to supply working R code that produces the results.
• You should copy-paste the R script (in the “source pane”) as text in the Appendix. For the R code
please avoid screenshots.
• Do not include substantive interpretation as comment in the script/appendix.
• You can briefly annotate your script (i.e., ## recoded variable to range from x to x, or ## producing
figure 2) and/or mark to which subquestion the code refers, although it is not mandatory.
Final remarks:
• While this is a take-home assignment and exam, hopefully you will find the topics engaging. Exams
can be overwhelming, but results emerge eventually.
Question 1: Sustainable food choices
You will be discussing the analysis and results presented in Hansen, P. G., Schilling, M., & Malthesen, M.
S. (2019). Nudging healthy and sustainable food choices: three randomized controlled field experiments
using a vegetarian lunch-default as a normative signal. Journal of Public Health.1 The first step will be to
carefully read the paper. All main analysis in this paper uses methodology that has been covered in your
coursework.
Question 1.1: Summarize the article’s main finding and highlight the contribution you deem most important and most surprising.
Question 1.2: Compare any two experiments of the paper in your own words: who are the treated units,
what is the response variable and what is the treatment? Discuss the role of randomization and the
differences in the treatment choices. Why did the authors rely on multiple experiments?
Question 1.3: Discuss considerations of internal and external validity throughout the whole study (all three experiments), highlighting potential trade-offs in general and how they relate to the studies carried out.
Question 1.4: The paper presents group differences in a plot and discusses these in text. Pick one comparison and carry out step-by-step the hypothesis testing (with the 𝛼 value of your choice) for a difference in proportion.
Question 1.5: The paper discusses some differences between men and women. Present a brief outline of
a study that you would design to get a more detailed view of these differences and the potential reasons
behind them.
Question 2: Personality traits and political attitudes
You will be working with parts of the data from Vitriol, J. A., E. G. Larsen & S. G. Ludeke. 2020. Just as WEIRD?
Personality Traits and Political Attitudes Among Immigrant Minorities. Journal of Research in Personality
85:2 Reading the paper can help you with giving better answers, however, the questions and tasks can be carried out without that.
You will have at your disposal a subset of some questions used in the paper (Immigrant panel (LISS-I
panel) in the Netherlands) in the traits.Rdata file (or traits.csv file). Not all respondents or items
are included in the data, thus your results will not be comparable to the ones in the paper. This is not a
reproduction exercise. The dataset is described below (see next page).
Before starting your work, you will have to take two decisions:
1. from the five Personality Trait variables (see table and/or paper), select one as your trait of core
interest. We will refer to this as “main predictor” in the tasks.
2. from Ideology, Political Trust, and Satisfaction with Democracy select one as your response/dependent
variable for the regression tasks. We will refer to this as “response/dependent” variable in the
tasks.
You do not have to give a reason why you chose what you chose, simply state your choice. You should
answer the following data analysis and interpretation related questions/tasks below. Please remember
the coding of the response/dependent variables.
1The article is also made available on Canvas.
2The article is also made available on Canvas. When you load the data, you should have 815 rows and 12 columns.
Question 2.1: Is the proportion of non-voters among first generation immigrants statistically significantly
different from 0.25? Carry out a hypothesis test step-by-step with your own calculations (you can use R for these) and also present the formulas you used. Use an 𝛼 level of 0.05. Do your conclusions change if you use an 𝛼 level of 0.1? Compare your results with the output from an R function that you find appropriate to answer this question. Please use two-tailed tests.
Question 2.2: Is the average level of agreeableness statistically significantly different between Dutch and
non-Dutch respondents? Present the hypothesis test either by carrying it out step-by-step or by relying
on a built in function for mean differences. Use an 𝛼 level of 0.05. Do your conclusions change if you use
an 𝛼 level of 0.1? Please use two-tailed tests.
Question 2.3: Summarize your main predictor (chosen above) variable using the appropriate measures
of central tendency and spread. Create a univariate summary plot and interpret what it displays. Create
a plot that summarizes the bivariate relationship between the response/dependent variable and the
chosen main predictor. Summarize the relationship through a correlation coefficient.
Question 2.4: For your response/dependent variable, calculate the standard error of the mean, construct
a 95% confidence interval for the mean and interpret these two uncertainty measures (standard error and confidence interval). Do this relying on your own calculations (you can use R for these) and also present the formulas you used.
Question 2.5: Run a linear regression where your chosen response/dependent variable is a function of
all five personality traits and the dutch, fg, and nw variables. Interpret the model results in substantive
terms and discuss measures of uncertainty and hypothesis test for the coefficients of interest.
Question 2.6: Extend the previous model in a way that can answer if there are heterogeneous effects on
your response/dependent variable regarding your chosen main predictor and non-Western background.
Report the model results in a table format and interpret the terms of interest and what this means for
your substantive conclusions. It is recommended, but not mandatory to offer a visual display of the
relationship.
For most of this question you have to carry out the analysis in R. This code has to be included as Appendix
and should produce the results featuring in your report. If something is reported in the text (also
figures/tables used), it has to be included in the code.
Question 3: Polls and election results in 2015 and 2019
We have looked at how well polls or betting markets predict electoral outcomes in the US. In the
poll-dk-comp.Rdata (or poll-dk-comp.csv) file you have Danish political party support polling numbers.
This dataset covers two parties, three polling companies, and two pre-election periods (2015 and 2019).
Using the concepts and metrics discussed in the course, analyzing solely the data at hand and relying
on your own analysis, answer the following questions. Please always describe how did you reach your
conclusion and what data/metric supports the conclusion. Depending on the measures you consider or
angle you take in your analysis, there might be different answers to these questions. This is normal, but
this is also why it is essential to describe why you chose those measures, how you made the comparisons
and how they substantiate your conclusion.
Question 3.1: In which pre-election did YouGov perform better, 2015 or 2019? How about Voxmeter?
Question 3.2: Relying on polls from both companies carried out in the last week before the elections (both 2015 and 2019), which party’s electoral result was better predicted?
Question 3.3: Relying on all polls, how did the quality of prediction change as we got closer to the 2015
election date? Please use at least one visualization to substantiate your conclusions. How was this different in 2019? Discuss the potential differences or similarities.
Question 3.4 Does the poll’s sample size matter for the quality of predictions, and if so, are there betweenparty differences in how important these are? Discuss the potential role of the timing of polls as well,relating your answer to point 3.3.
For most of this question you have to carry out the analysis in R. This code has to be included as Appendix and should produce the results featuring in your report. If something is reported in the text (also figures/tables used), it has to be included in the code.
