Gov1372 -Data Exploration - Emotional Arousal - Solved

In this Data Exploration assignment we will explore Clifford and Jerit’s (2018) findings about the effects of disgust and anxiety on political learning.

If you have a question about any part of this assignment, please ask! Note that the actionable part of each question is bolded.

Emotional Arousal: Disgust and Anxiety
Data Details:

•     File Name: Study1ReplicationData.dta

•     Source: These data are from Study 1 in Clifford and Jerit (2018).

 

Variable Name
Variable Description

 
treat_rand1
Treatment assignment: 1-Low Anxiety/Low Disgust,

2-High Anxiety/Low Disgust, 3-Low Anxiety/High Disgust, and 4-High Anxiety/High Disgust
Q11_1
Self reported feeling of how well DISGUST describes respondent’s emotional reaction to virus: 1-Not Well At

All, 2- Not Too Well, 3-Somewhat Well, 4-Very Well, 5-Extremely Well, 8-Skipped
Q11_2
Self reported feeling of how well GROSSED OUT describes respondent’s emotional reaction to virus:

1-Not Well At All, 2- Not Too Well, 3-Somewhat Well, 4-Very Well, 5-Extremely Well, 8-Skipped
Q11_3
Self reported feeling of how well REPULSED describes respondent’s emotional reaction to virus: 1-Not Well At

All, 2- Not Too Well, 3-Somewhat Well, 4-Very Well, 5-Extremely Well, 8-Skipped
Q11_4
Self reported feeling of how well AFRAID describes respondent’s emotional reaction to virus: 1-Not Well At

All, 2- Not Too Well, 3-Somewhat Well, 4-Very Well, 5-Extremely Well, 8-Skipped
Q11_5
Self reported feeling of how well ANXIOUS describes respondent’s emotional reaction to virus: 1-Not Well At

All, 2- Not Too Well, 3-Somewhat Well, 4-Very Well, 5-Extremely Well, 8-Skipped
Q11_6
Self reported feeling of how well WORRIED describes respondent’s emotional reaction to virus: 1-Not Well At

All, 2- Not Too Well, 3-Somewhat Well, 4-Very Well, 5-Extremely Well, 8-Skipped
 

Variable Name
Variable Description

 
Q12_1
Identification of FATIGUE as a symptom: 1-Yes, 2- No
Q12_2
Identification of HEADACHES as a symptom: 1-Yes, 2-

No
Q12_3
Identification of DIARRHEA as a symptom: 1-Yes, 2-

No
Q12_4
Identification of JOINT PAIN as a symptom: 1-Yes, 2-

No
Q12_5
Identification of BOILS as a symptom: 1-Yes, 2- No
Q12_6
Identification of WARTS as a symptom: 1-Yes, 2- No
Q12_7
Identification of FEVER as a symptom: 1-Yes, 2- No
Q13
Identification of method of disease transmission:

1-Person to Person Contact, 2- The air, 3- Animals, 4-Insects, 5-Food, 8-Skipped
Q14
Identification of there being a cure for the virus: 1-Yes,

2-No, 8-Skipped
Q15
Requested additional information be sent to them:

1-Yes, 2-No, 8-Skipped
Q16_1
Topic of requested information AFFECTED

COUNTRIES: 1-Yes, 2-No, 9-Not Asked
Q16_2
Topic of requested information TRV IN US: 1-Yes,

2-No, 9-Not Asked
Q16_3
Topic of requested information TRV TRANSMISSION:

1-Yes, 2-No, 9-Not Asked
Q16_4
Topic of requested information AT RISK

POPULATION: 1-Yes, 2-No, 9-Not Asked
Q16_5
Topic of requested information DEATH TOLL: 1-Yes,

2-No, 9-Not Asked
Q16_6
Topic of requested information PROGRESS ON CURE:

1-Yes, 2-No, 9-Not Asked
Q16_7
Topic of requested information SYMPTOMS: 1-Yes,

2-No, 9-Not Asked
Q17_6
Self-reported likelihood of looking up more info: 1-Not likely at all, 2-Not too likely, 3-Somewhat likely, 4-Very likely, 5-Extremely likely, 8-Skipped
Q17_7
Self-reported likelihood of talking with friends or family about disease in next week: 1-Not likely at all, 2-Not too likely, 3-Somewhat likely, 4-Very likely, 5-Extremely likely, 8-Skipped
page_article_timing
Time spent in seconds on page containing article about disease
Q18
Self-reported gender: 1-Female, 0-Not Female
Q19
Self-reported race/ethnicity: 1-White, 2-Black,

3-Hispanic, 4-Asian, 5-Native American, 6-Mixed Race, 7-Other
Q20
Self-reported education: 1-No HS, 2-High school graduate, 3-Some college, 4-2 year degree, 5-4 year degree, 6-Post-graduate degree
Q21
Self-reported partisanship: 1-Strong Democrat, 2-Not very strong Democrat, 3-Lean Democrat,

4-Independent, 5-Lean Republican, 6-Not very strong Republican, 7- Strong Republican, 8-Not Sure
Variable Name
Variable Description
 
Q22
Self-reported voter registration status: 1-Yes, 2-No,

3-Don’t know
 
Q23
Self-reported ideology: 1-Very Liberal, 2-Liberal, 3-Moderate, 4-Conservative, 5-Very Conservative
 
#Load the data for Study 1
Study1_preprocess <- read_dta("Study1ReplicationData.dta")

To date you have read in data from both .csv and .RData files. The data for this week are stored in another common file type, .dta files. This is the format for data exported using Stata, the other most commonly used statistical software package besides R. Reading these files requires the read_dta() function in the haven package, so be sure to install it if you have not already!

Question 1 DO NOT SKIP
Cleaning and organizing data is an important part of any research process. Note: Your blog posts should not address the data cleaning portion of the assignment but rather the content of the material

Part a

The data is not currently in the most usable condition. As currently read in, many of the dataset’s variables are somewhat unhelpfully labelled with just the survey question number. In order to make them more intuitive to work with we can rename them easily using the dplyr package. Here we will make use of the rename_with() function. The .cols argument in the rename_with() function specifies which columns should be renamed, which takes either the original name of the variable, the index, or a logical argument. We will explore each of these methods beginning with renaming variables by name. Modify the code below to rename variables Q13, Q14, and Q15. Make sure the list of replacement variable names has the same number as the number of variables you’re renaming. Also be sure to save the modified dataset as a new object to continue working with it.

#Method 1: Renaming by name
Study1_processing1 <- Study1_preprocess %>% rename_with(.cols = c(Q13, Q14, Q15), ~c("PUT COMMA SEPARATED LIST OF VARIABLE NAMES AS STRINGS HERE F

## Error: Names must be unique.

## x These names are duplicated:

##                       * "PUT COMMA SEPARATED LIST OF VARIABLE NAMES AS STRINGS HERE FOR THESE THREE VARIABLES IN ORDER" a

Part b

Many other variables are also unhelpfully labelled. For example the last six variables in the data are demographic characteristics (gender, race, education, party id, voter registration status, and ideology). You can also rename variables by position using the following code. Modify the code below to rename the demographic variables. Make sure the list of replacement variable names has the same number as the number of variables you’re renaming. Also be sure to save the modified dataset as a new object to continue working with it.

#Method 2: Renaming by index/position

#last_col() is just a function that returns the index number of the last variable in the dataset (33 in

Study1_processing2 <- Study1_processing1 %>% rename_with(.cols = last_col(offset = 5):last_col(), ~c(
"PUT COMMA SEPARATED LIST OF VARIABLE NAMES AS

## Error in rename_with(., .cols = last_col(offset = 5):last_col(), ~c("PUT COMMA SEPARATED LIST OF VARI

Part c

We can also rename all variables that satisfy a logical condition. Note that every variable relating to emotional reaction of the respondent is labelled as a part of question 11 (Q11). Modify the code below to rename the demographic variables. Make sure the list of replacement variable names has the same number as the number of variables you’re renaming. Also be sure to save the modified dataset as a new object to continue working with it.

#Method 3: Renaming by logical condition
Study1_processing3 <- Study1_processing2 %>% rename_with(.cols = contains("Q11"), ~c("PUT COMMA SEPARATED LIST OF EMOTIONS NAMES AS STRINGS HERE FO

## Error in rename_with(., .cols = contains("Q11"), ~c("PUT COMMA SEPARATED LIST OF EMOTIONS NAMES AS ST

Part d

We may not need all the variables in the dataset. For instance, the analysis below will not rely on knowing what topic of additional information people were interested in (Q16_1 through Q16_7). The code below will search for a string in the variable labels and return only the variables that DO NOT include that string. Modify the code and use it to drop the unneeded variables from the dataset. Be sure to save the modified dataset as a new object to continue working with it.

Study1_processing4 <- Study1_processing3 %>% select(-contains("TARGETSTRING"))

## Error in select(., -contains("TARGETSTRING")): object 'Study1_processing3' not found

Part e

Another issue often encountered is that the way missing data is coded can vary across data sources. Here responses are marked with the value of 8 for the respondent skipping the question. The functions mutate() and across() used in tandem are useful for recoding many variables at once in the same manner. across() specifies which columns to mutate and what function to apply to them all, in this case we’ll want to use the function na_if() which recodes a variable as NA if it matches the second argument. Modify the code below to change the relevant data columns to be binary variables with NA for missing data.

Hint: For this dataset all values of 8 indicate a kind of missing data.

Study1_processing5 <- Study1_processing4 %>% mutate(across(.cols = "PUT TARGET COLUMNS HERE", ~na_if(.,"PUT TARGET VALUE HERE")))

## Error in mutate(., across(.cols = "PUT TARGET COLUMNS HERE", ~na_if(., : object 'Study1_processing4'

Part f

Some of the binary data are also coded with 1s and 2s as opposed to 0s and 1s. These include the identification of symptoms, the request for additional information, identification that there is a cure, and voter registration status. Use the mutate function to recode those variables to be binary with 1 for yes and 0 for no. Hint: You can look at the solutions for previous data exercises but you’ll want to combine the mutate() and ifelse() functions.

Part g

Lastly we may be interested not just in individual symptom recall but also whether or not the respondent correctly remembered all the symptoms in their treatment with no mistakes. Use the mutate() function to add a variable capturing whether or not the respondent correctly identified the symptoms in their treatment. Remember that which set of symptoms are correct differs somewhat by treatment! Warning: This one may be a bit tricky, no worries if you don’t quite get it. Feel free to skip to the next question.

Part h

Use the tools above to alter the data in whatever way you see fit. Some examples could be renaming remaining variables, creating new binary variables that identify if respondents are part of a certain racial or ethnic group (these would be necessary for including race in a regression for example), or any other transformation of the data. Use the tools above or any others to modify the data to more useful for the following exercises in any way you see fit. Save this final version as the object you’ll use for future questions.

Question 2
Part a

What are the treatments in Study 1? How many treatment conditions are there in Study 1? What are they? How many respondents were in each condition? Hint: Look at page 269 for the treatment conditions of Study 1

Part b

The paper lays out three distinct hypotheses concerning the impact of disgust on information uptake and search. What are the three hypotheses? Which outcome variables in Study 1 speak to which of these hypotheses? Hint: Look at pages 267 and 268 for the hypotheses

Question 3
Part a

Often times when running an experiment a researcher will include a ‘manipulation check’ to confirm that the treatment was noticed and is having some of its intended effect. In this study they ask about the emotional response to the fictitious virus using two three item emotional indices: anxiety (afraid, anxious, worried) and disgust (disgusted, grossed out, repulsed). Take the average of the responses for each emotional index and check to see if the treatments had the desired manipulation. Compare the average anxiety and disgust responses for each of the four treatments. Are there statistically significant differences in the way one would expect?

Part b

Choose one of the three hypotheses identified in Question 2. Compare the responses on one or more of the outcomes relevant to that hypothesis. Is the hypothesis supported by the data? For one or more outcome variables relevant to one of the three hypotheses check for statistically significant differences. What can we conclude from the data?

Question 4: DATA SCIENCE QUESTION
Part a

As we have spoken about in class, using indexes of multiple measures aimed at a single concept is often more reliable than using only one measure. However how is one to know that all the measures in the index are related to the same concept? Cronbach’s alpha is a measure of how internally consistent the answers to multiple questions are. It is given by the formula:

N ×Pc α =  

Pv + (N − 1) ×Pc

Where N is the number of items in the index, Pc is the sum of the covariances for item pairs, and Pv is the sum of the variance for the items. Using the formula calculate Cronbach’s alpha for the disgust index and the anxiety indexes used in the paper.

Part b

In R, one can use the function cronbach.alpha() from the ltm package. Generally scales or indices with a Cronbach’s alpha below 0.7 are considered insufficiently internally consistent for use. Calculate Cronbach’s alpha for the disgust index and the anxiety indexes using the function in R. Does this match your previous calculation? Are each of the scales sufficiently internally consistent?

install.packages("ltm")

## Error in contrib.url(repos, "source"): trying to use CRAN without setting a mirror

library(ltm)

## Loading required package: MASS

##

## Attaching package: 'MASS'

## The following object is masked _by_ '.GlobalEnv':

##

##              select

## The following object is masked from 'package:dplyr':

##

##              select

## Loading required package: msm

## Loading required package: polycor

Part c

Create a multivariate regression model for an outcome variable of your choosing. Carefully interpret the results. Create a multivariate linear regression model for any outcome of your choosing with any covariates of your choice. Justify your choice of models and what the result may indicate. Carefully interpret the coefficients for the model.

Part d

You can create coefficient plots using regression coefficients as well. Use the results of part c to create a plot of the coefficients in your regression model. Create a plot of the regression coefficients with their 95% confidence intervals. Be sure to include a line demarcating 0. Hint: the confint() function can take a regression object and return the upper and lower bounds of the confidence intervals.

Emotional Arousal: Disgust
Data Details:

•     File Name: Study2ReplicationData.dta

•     Source: These data are from Study 2 in Clifford and Jerit (2018).

 

Variable Name
Variable Description

 
treament
Treatment assignment: 1-Control, 2-Disgusting

Imagery/No Map, 3-Map/No Disgusting Imagery, and

4-Disgusting Imagery and Map
page_time
Time spent viewing page with treatment article
symptpercent
Belief about percentage of people who contract Dengue Fever but never experience symptoms: 1-0%, 2-20%, 3-40%, 4-60%, 5-80%
Mexico
Identify MEXICO to be at risk for spread of Dengue:

0-No, 1-Yes
SouthAmerica
Identify SOUTH AMERICA to be at risk for spread of

Dengue: 0-No, 1-Yes
Africa
Identify AFRICA to be at risk for spread of Dengue:

0-No, 1-Yes
Canada
Identify CANADA to be at risk for spread of Dengue:

0-No, 1-Yes
Russia
Identify RUSSIA to be at risk for spread of Dengue:

0-No, 1-Yes
Europe
Identify EUROPE to be at risk for spread of Dengue:

0-No, 1-Yes
length
Belief about how long symptoms of Dengue Fever typically last: 1-A few days, 2-A week, 3-Two to three weeks, 4-A month or more
fever
Identification of FEVER as a symptom: 0-No, 1-Yes
headache
Identification of HEADACHE as a symptom: 0-No,

1-Yes
jointpain
Identification of JOINT PAIN as a symptom: 0-No,

1-Yes
rash
Identification of RASH as a symptom: 0-No, 1-Yes
bleeding
Identification of BLEEDING FROM EYES, NOSE,

AND GUMS as a symptom: 0-No, 1-Yes
nausea
Identification of NAUSEA as a symptom: 0-No, 1-Yes
seizures
Identification of SEIZURES as a symptom: 0-No, 1-Yes
breathing
Identification of DIFFICULTY BREATHING as a symptom: 0-No, 1-Yes
infosearch
Self-reported likelihood of looking up more info: 1-Not likely at all, 2-Not too likely, 3-Somewhat likely, 4-Very likely, 5-Extremely likely
talk
Self-reported likelihood of talking with friends or family about disease in next week: 1-Not likely at all, 2-Not too likely, 3-Somewhat likely, 4-Very likely, 5-Extremely likely
 

 
Variable Name
Variable Description
 
 
infosession
Request invitation to information session about Dengue

Fever: 0-No, 1-Yes
 
 
learn
Request additional info about Dengue Fever in survey:

0-No, 1-Yes
 
 
infogiveemail
Provided email address to receive invitation to info session: 0-No, 1-Yes
 
 
E_disgust
Self reported feeling of how well DISGUSTED describes respondent’s emotional reaction to disease:

1-Not Well At All, 2- Not Too Well, 3-Somewhat Well, 4-Very Well, 5-Extremely Well
 
 
E_gross
Self reported feeling of how well GROSSED OUT describes respondent’s emotional reaction to disease:

1-Not Well At All, 2- Not Too Well, 3-Somewhat Well, 4-Very Well, 5-Extremely Well
 
 
E_resentment
Self reported feeling of how well RESENTFUL describes respondent’s emotional reaction to disease:

1-Not Well At All, 2- Not Too Well, 3-Somewhat Well, 4-Very Well, 5-Extremely Well
 
 
E_revulsion
Self reported feeling of how well REVULSION describes respondent’s emotional reaction to disease:

1-Not Well At All, 2- Not Too Well, 3-Somewhat Well, 4-Very Well, 5-Extremely Well
 
 
E_hateful
Self reported feeling of how well HATEFUL describes respondent’s emotional reaction to disease: 1-Not Well

At All, 2- Not Too Well, 3-Somewhat Well, 4-Very Well,

5-Extremely Well
 
 
E_angry
Self reported feeling of how well ANGRY describes respondent’s emotional reaction to disease: 1-Not Well

At All, 2- Not Too Well, 3-Somewhat Well, 4-Very Well,

5-Extremely Well
 
 
E_anxiety
Self reported feeling of how well ANXIETY describes respondent’s emotional reaction to disease: 1-Not Well

At All, 2- Not Too Well, 3-Somewhat Well, 4-Very Well,

5-Extremely Well
 
 
E_nervous
Self reported feeling of how well NERVOUS describes respondent’s emotional reaction to disease: 1-Not Well

At All, 2- Not Too Well, 3-Somewhat Well, 4-Very Well,

5-Extremely Well
 
 
E_worry
Self reported feeling of how well WORRIED describes respondent’s emotional reaction to disease: 1-Not Well

At All, 2- Not Too Well, 3-Somewhat Well, 4-Very Well,

5-Extremely Well
 
Study2_preprocess <- read_dta("Study2ReplicationData.dta")

Study2 <- Study2_preprocess %>%

#grouping similar treatment conditions and computing time spent viewing treatment mutate(treatment = case_when(c_control==1 ~ 1, c_bothd==1|c_bothm==1 ~ 4, c_disguste==1|c_disgustl==1 ~ 2, c_mapl==1|c_mape==1 ~3),
 
 
 
 
page_time = rowMeans(select(., contains("t_c")), na.rm = TRUE)) %>%

#recoding NAs as 0 for country and symptom variables mutate(across(contains(c("countries","symptoms")), ~ifelse(is.na(.x),0,.x))) %>%

#recoding infosession as binary mutate(across(c(infosession,learn), ~ifelse(.x==1,1,0))) %>%

#renaming country variables rename_with(.cols= contains("countries"), ~c("Mexico",

"SouthAmerica", "Africa", "Canada", "Russia",

"Europe")) %>% #renaming symptom variables rename_with(.cols = contains("ksymptoms"), ~c("fever",

"headache",

"jointpain", "rash",

"bleeding", "nausea",

"seizures", "breathing")) %>%

#renaming info search variables and other disease knowledge info rename_with(.cols = contains("search"), ~c("infosearch",

"talk")) %>%

rename(symptpercent = kpercent, length = klength) %>%

#renaming emotion variables rename_with(.cols = contains("emotion"), ~paste0("E_", c("disgust",

"gross",

"resentment",

"revulsion",

"hateful", "angry",

"anxiety", "nervous",

"worry"))) %>% #creating indicator variable for correct identification of symptoms and at risk countries mutate(country_correct = as.numeric(paste0(Mexico, SouthAmerica, Africa, Canada, Russia, Europe symptoms_correct = as.numeric(paste0(

#deleting irrelevant variables select(-contains("c_")) %>%

#reordering treatment as first variable relocate(treatment)
))==111 fever, headache, jointpain, rash, bleeding, nausea, seizur

Question 5
Above is an example of the kind of data cleaning that sometimes must be done to make the datasets intuitive and easy to work with. Look through the code above and try to follow what each line is doing. Look at the Study1_preprocess and Study1 datasets and note the differences between them. You do not need to write anything for this question, just get a sense of some of the useful tools when cleaning data!

Question 6
Part a

What are the treatment conditions for Study 2? What are the treatment conditions for Study 2? How do they differ from Study 1? Hint: Look at page 273 for the treatment conditions of Study 2

Part b

Study 2 asks about three different categories of emotions: disgust (disgusted, grossed out, revulsion), anxiety (anxious, nervous, worried), and anger (angry, hateful, resentful). Did the treatments succeed in manipulating emotions? Was the impact limited to disgust? Pool respondents into low disgust and high disgust treatments. Check for statistically significant differences in the average answer to each emotion index.

Part c

Plot the distribution of disgust, anxiety, and anger for low disgust and high disgust treatment conditions. Pool respondents into low disgust and high disgust treatments. Plot the distributions for the average item response to each of the three emotional categories.

Question 7
Part a

The researchers tested whether the inclusion of any kind of imagery (the map) would affect informational recall. Does the map impact information recall about the info it shows (affected countries)? What about other information? Compare the accuracy of affected country recall in map and non-map treatment conditions. Then choose one other measure information recall and compare it across map and non-map treatments.

Part b

Choose one of the three hypotheses identified in Question 2. Compare the responses on one or more of the outcomes relevant to that hypothesis using data from Study 2. Is the hypothesis supported by the data? For one or more outcome variables relevant to one of the three hypotheses check for statistically significant differences. What can we conclude from the data?

Part c

One argument the paper acknowledges is that those in the disgust treatment may simply be clicking through the treatment quickly to avoid the imagery and this is affecting recall. Compare mean page_view_length for each group. Do the disgusting images cause people to spend less time viewing the page?

Question 8
Part a

Compare your intepretations of the results of both studies to the paper’s interpretation. What is your interpretation of the findings across both studies? Do your takeaways match or differ those of the authors?

Part b

What other emotions may be of interest? What other emotions may affect information uptake or political behavior? How so? How might you test these hypotheses?

Question 9
Run a linear regression on an outcome variable of interest (e.g. searching for more information or correctly identifying all symptoms) using any of the variables in the dataset for either Study 1 or Study 2. Run an OLS model for any outcome variable of interest with your own specification. Carefully interpret the results. What do they tell us?