Rebecca Gibble Final Project
Analysis Overview
I am interested in analyzing how those who graduated high school only compare to those who graduated from college (variable: educ_2019, which is the indepdent variable) in their feelings about women and their role in society. I will be analyzing the following dependent variables:
- Gender equality issue importance (imiss_y_2019)
- Gender roles (sexism1_2019)
- Women should return to their traditional roles in society
- Modern sexism demands (sexism2_2019)
- Women often miss out on good jobs because of discrimination
- Feeling towards women on a scale from 1-100 (Women_2019)
*As a note, when referring to college graduates throughout this analysis, that entails those who graduated from a four year institution.
Importing the data
library(dplyr)## Warning: package 'dplyr' was built under R version 3.6.2##
## Attaching package: 'dplyr'## The following objects are masked from 'package:stats':
##
## filter, lag## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, unionlibrary(readr)## Warning: package 'readr' was built under R version 3.6.2library(ggplot2)## Warning: package 'ggplot2' was built under R version 3.6.2data<-read_csv("/Users/rebeccagibble/Downloads/Voter Data 2019 (1).csv")##
## ── Column specification ────────────────────────────────────────────────────────
## cols(
## .default = col_double(),
## weight_18_24_2018 = col_logical(),
## izip_2019 = col_character(),
## housevote_other_2019 = col_character(),
## senatevote_other_2019 = col_character(),
## senatevote2_other_2019 = col_character(),
## SenCand1Name_2019 = col_character(),
## SenCand1Party_2019 = col_character(),
## SenCand2Name_2019 = col_character(),
## SenCand2Party_2019 = col_character(),
## SenCand3Name_2019 = col_character(),
## SenCand3Party_2019 = col_character(),
## SenCand1Name2_2019 = col_character(),
## SenCand1Party2_2019 = col_character(),
## SenCand2Name2_2019 = col_character(),
## SenCand2Party2_2019 = col_character(),
## SenCand3Name2_2019 = col_character(),
## SenCand3Party2_2019 = col_character(),
## governorvote_other_2019 = col_character(),
## GovCand1Name_2019 = col_character(),
## GovCand1Party_2019 = col_character()
## # ... with 108 more columns
## )
## ℹ Use `spec()` for the full column specifications.## Warning: 800 parsing failures.
## row col expected actual file
## 2033 weight_18_24_2018 1/0/T/F/TRUE/FALSE .917710168467982 '/Users/rebeccagibble/Downloads/Voter Data 2019 (1).csv'
## 2828 weight_18_24_2018 1/0/T/F/TRUE/FALSE 1.41022291345592 '/Users/rebeccagibble/Downloads/Voter Data 2019 (1).csv'
## 4511 weight_18_24_2018 1/0/T/F/TRUE/FALSE 1.77501243840922 '/Users/rebeccagibble/Downloads/Voter Data 2019 (1).csv'
## 7264 weight_18_24_2018 1/0/T/F/TRUE/FALSE 1.29486870319614 '/Users/rebeccagibble/Downloads/Voter Data 2019 (1).csv'
## 7277 weight_18_24_2018 1/0/T/F/TRUE/FALSE 1.44972719707603 '/Users/rebeccagibble/Downloads/Voter Data 2019 (1).csv'
## .... ................. .................. ................ ........................................................
## See problems(...) for more details.head(data)## # A tibble: 6 x 1,282
## weight_2016 weight_2017 weight_panel_20… weight_latino_2… weight_18_24_20…
## <dbl> <dbl> <dbl> <dbl> <lgl>
## 1 0.358 0.438 0.503 NA NA
## 2 0.563 0.366 0.389 NA NA
## 3 0.552 0.550 0.684 NA NA
## 4 0.208 NA NA NA NA
## 5 0.334 0.346 0.322 NA NA
## 6 0.207 0.148 0.594 NA NA
## # … with 1,277 more variables: weight_overall_2018 <dbl>, weight_2019 <dbl>,
## # weight1_2018 <dbl>, weight1_2019 <dbl>, weight2_2019 <dbl>,
## # weight3_2019 <dbl>, cassfullcd <dbl>, vote2020_2019 <dbl>,
## # trumpapp_2019 <dbl>, fav_trump_2019 <dbl>, fav_obama_2019 <dbl>,
## # fav_hrc_2019 <dbl>, fav_sanders_2019 <dbl>, fav_putin_2019 <dbl>,
## # fav_schumer_2019 <dbl>, fav_pelosi_2019 <dbl>, fav_comey_2019 <dbl>,
## # fav_mueller_2019 <dbl>, fav_mcconnell_2019 <dbl>, fav_kavanaugh_2019 <dbl>,
## # fav_biden_2019 <dbl>, fav_warren_2019 <dbl>, fav_harris_2019 <dbl>,
## # fav_gillibrand_2019 <dbl>, fav_patrick_2019 <dbl>, fav_booker_2019 <dbl>,
## # fav_garcetti_2019 <dbl>, fav_klobuchar_2019 <dbl>, fav_gorsuch_2019 <dbl>,
## # fav_kasich_2019 <dbl>, fav_haley_2019 <dbl>, fav_bloomberg_2019 <dbl>,
## # fav_holder_2019 <dbl>, fav_avenatti_2019 <dbl>, fav_castro_2019 <dbl>,
## # fav_landrieu_2019 <dbl>, fav_orourke_2019 <dbl>,
## # fav_hickenlooper_2019 <dbl>, fav_pence_2019 <dbl>, add_confirm_2019 <dbl>,
## # izip_2019 <chr>, votereg_2019 <dbl>, votereg_f_2019 <dbl>,
## # regzip_2019 <dbl>, region_2019 <dbl>, turnout18post_2019 <dbl>,
## # tsmart_G2018_2019 <dbl>, tsmart_G2018_vote_type_2019 <dbl>,
## # tsmart_P2018_2019 <dbl>, tsmart_P2018_party_2019 <dbl>,
## # tsmart_P2018_vote_type_2019 <dbl>, housevote_2019 <dbl>,
## # housevote_other_2019 <chr>, senatevote_2019 <dbl>,
## # senatevote_other_2019 <chr>, senatevote2_2019 <dbl>,
## # senatevote2_other_2019 <chr>, SenCand1Name_2019 <chr>,
## # SenCand1Party_2019 <chr>, SenCand2Name_2019 <chr>,
## # SenCand2Party_2019 <chr>, SenCand3Name_2019 <chr>,
## # SenCand3Party_2019 <chr>, SenCand1Name2_2019 <chr>,
## # SenCand1Party2_2019 <chr>, SenCand2Name2_2019 <chr>,
## # SenCand2Party2_2019 <chr>, SenCand3Name2_2019 <chr>,
## # SenCand3Party2_2019 <chr>, governorvote_2019 <dbl>,
## # governorvote_other_2019 <chr>, GovCand1Name_2019 <chr>,
## # GovCand1Party_2019 <chr>, GovCand2Name_2019 <chr>,
## # GovCand2Party_2019 <chr>, GovCand3Name_2019 <chr>,
## # GovCand3Party_2019 <chr>, inst_court_2019 <dbl>, inst_media_2019 <dbl>,
## # inst_congress_2019 <dbl>, inst_justice_2019 <dbl>, inst_FBI_2019 <dbl>,
## # inst_military_2019 <dbl>, inst_church_2019 <dbl>, inst_business_2019 <dbl>,
## # Democrats_2019 <dbl>, Republicans_2019 <dbl>, Men_2019 <dbl>,
## # Women_2019 <dbl>, wm_2019 <dbl>, ww_2019 <dbl>, bm_2019 <dbl>,
## # bw_2019 <dbl>, hm_2019 <dbl>, hw_2019 <dbl>, rwm_2019 <dbl>,
## # rww_2019 <dbl>, rbm_2019 <dbl>, rbw_2019 <dbl>, pwm_2019 <dbl>, …Recoding and selecting variables
Recoding variables from numeric form to their labeled form.
Selecting only necessary variables.
newdata<-data%>%
mutate(GenderEquality = ifelse(imiss_y_2019==1,"Very Important",
ifelse(imiss_y_2019==2,"Somewhat Important",
ifelse(imiss_y_2019==3, "Not very Important",
ifelse(imiss_y_2019==4, "Unimportant", NA)))),
GenderRoles = ifelse(sexism1_2019==1, "Strongly Agree",
ifelse(sexism1_2019==2, "Somewhat Agee",
ifelse(sexism1_2019==3, "Somewhat Disagree",
ifelse(sexism1_2019==4, "Strongly Disagree",NA)))),
ModernSexism = ifelse(sexism2_2019==1, "Strongly Agree",
ifelse(sexism2_2019==2, "Somewhat Agree",
ifelse(sexism2_2019==3, "Somewhat Disagree",
ifelse(sexism2_2019==4, "Strongly Disagree",NA)))),
FeelingAboutWomen = ifelse(Women_2019>100,NA, Women_2019))%>%
select(GenderEquality,GenderRoles,ModernSexism,FeelingAboutWomen,educ_2019)
head(newdata)## # A tibble: 6 x 5
## GenderEquality GenderRoles ModernSexism FeelingAboutWom… educ_2019
## <chr> <chr> <chr> <dbl> <dbl>
## 1 Very Important Strongly Disagr… Strongly Disagr… 80 5
## 2 <NA> <NA> <NA> NA NA
## 3 Very Important Strongly Disagr… Somewhat Disagr… 71 2
## 4 Not very Importa… Strongly Disagr… Strongly Agree 10 3
## 5 Somewhat Importa… Strongly Disagr… Somewhat Disagr… 95 5
## 6 Very Important Strongly Disagr… Strongly Disagr… 100 4Filtering Data
Filtering data to only select two groups needed for analysis.
Previewing data to confirm changes.
newdata2<-newdata%>%
filter(educ_2019 %in% c("2","5"))%>%
mutate(EducationLevel = ifelse(educ_2019==2, "High school graduate",
ifelse(educ_2019==5, "College graduate",NA)))%>%
select(GenderEquality,GenderRoles,ModernSexism,FeelingAboutWomen,EducationLevel)
head(newdata2)## # A tibble: 6 x 5
## GenderEquality GenderRoles ModernSexism FeelingAboutWom… EducationLevel
## <chr> <chr> <chr> <dbl> <chr>
## 1 Very Important Strongly Dis… Strongly Disa… 80 College gradua…
## 2 Very Important Strongly Dis… Somewhat Disa… 71 High school gr…
## 3 Somewhat Import… Strongly Dis… Somewhat Disa… 95 College gradua…
## 4 Very Important Strongly Dis… Strongly Disa… 99 College gradua…
## 5 Very Important Strongly Dis… Strongly Disa… 82 College gradua…
## 6 Very Important Strongly Dis… Strongly Disa… 99 College gradua…Analyses for each dependent variable vs. the independent variable
Analysis 1
Analyzing relationship between education level (EducationLevel) and gender equality issue importance (GenderEquality).
Crosstab: Gender Equality
table(newdata2$GenderEquality,newdata2$EducationLevel)%>%
prop.table(2)##
## College graduate High school graduate
## Not very Important 0.1607247 0.1713198
## Somewhat Important 0.2916423 0.3362944
## Unimportant 0.1455289 0.1472081
## Very Important 0.4021040 0.3451777From this crosstab we can see that college educated respondents are more likely to find that gender equality is very important compared to only high school graduates.
Stacked Bar Chart: Gender Equality
newdata2%>%
filter(!is.na(GenderEquality))%>%
group_by(EducationLevel,GenderEquality)%>%
summarize(n=n())%>%
mutate(percent=n/sum(n))%>%
ggplot()+
geom_col(aes(x=EducationLevel,y=percent,fill=GenderEquality))+
theme_minimal()## `summarise()` regrouping output by 'EducationLevel' (override with `.groups` argument)
The stacked barchart illustrates the same data as noted above while simultaneously making it easier to tell the difference between the different categories. On average, college graduates are more likely to believe that gender equality is very important which we can see from the area of the purple bar for college graduates being greater than the area of the purple bar for high school graduates.
Chi-Square Test: Gender Equality
chisq.test(newdata2$GenderEquality,newdata2$EducationLevel)##
## Pearson's Chi-squared test
##
## data: newdata2$GenderEquality and newdata2$EducationLevel
## X-squared = 12.889, df = 3, p-value = 0.004883There is a statistically significant relationship between education level and the importance of gender equality. We know this because the p-value obtained from the chi-squared test of 0.004883 is less than 0.05.
Analysis 2
Analyzing relationship between education level (EducationLevel) and gender roles (women should return to their traditional roles in society) (variable:GenderRoles).
Crosstab: Gender Roles
table(newdata2$GenderRoles,newdata2$EducationLevel)%>%
prop.table(2)##
## College graduate High school graduate
## Somewhat Agee 0.10599884 0.16379860
## Somewhat Disagree 0.22947001 0.29063098
## Strongly Agree 0.03669190 0.07329509
## Strongly Disagree 0.62783925 0.47227533College educated respondents are approximately 1.3 times more likely to strongly disagree with women returning to their taditional roles in society versus high school graduates.
Stacked Bar Chart: Gender Roles
newdata2%>%
filter(!is.na(GenderRoles))%>%
group_by(EducationLevel,GenderRoles)%>%
summarize(n=n())%>%
mutate(percent=n/sum(n))%>%
ggplot()+
geom_col(aes(x=EducationLevel,y=percent,fill=GenderRoles))+
theme_minimal()## `summarise()` regrouping output by 'EducationLevel' (override with `.groups` argument)
The stacked barchart overall shows that college graduates, in general, are more likely to think that women are not obligated to return to their traditional gender roles in society. We see this in the “Strongly Disagree” category (purple bar) which is greater for college graduates than high school graduates.
Chi-Square Test: Gender Roles
chisq.test(newdata2$GenderRoles,newdata2$EducationLevel)##
## Pearson's Chi-squared test
##
## data: newdata2$GenderRoles and newdata2$EducationLevel
## X-squared = 88.475, df = 3, p-value < 2.2e-16There is a statistically significant relationship between education level and women returning to their traditional gender roles in society. We know this because the p-value obtained from this chi-squared test of 2.2e-16 is less than 0.05. In fact, the relationship between education level and gender roles is more significant than the relationship between education level and gender equality (analyzed above). This is known because the p-value in this analysis (2.2e-16) is smaller than the p-value in Analysis 1 (0.004883).
Analysis 3
Analyzing relationship between education level (EducationLevel) and modern sexism demands (women often miss out on good jobs because of discrimination) (variable:ModernSexism).
Crosstab: Modern Sexism
table(newdata2$ModernSexism,newdata2$EducationLevel)%>%
prop.table(2)##
## College graduate High school graduate
## Somewhat Agree 0.2216405 0.2305236
## Somewhat Disagree 0.2082606 0.2688378
## Strongly Agree 0.1436882 0.1819923
## Strongly Disagree 0.4264107 0.3186462High school graduates are more likely to strongly agree that women often miss out on good jobs due to discrimination as compared to college graduates, however, not by much. The greatest difference seen above is that college graduates are more likely to strongly disagree that women often miss out on good jobs due to discrimination.
Stacked Bar Chart: Modern Sexism
newdata2%>%
filter(!is.na(ModernSexism))%>%
group_by(EducationLevel,ModernSexism)%>%
summarize(n=n())%>%
mutate(percent=n/sum(n))%>%
ggplot()+
geom_col(aes(x=EducationLevel,y=percent,fill=ModernSexism))+
theme_minimal()## `summarise()` regrouping output by 'EducationLevel' (override with `.groups` argument)
The stacked barchart illustrates more distinctly the differences in opinion between college and high school graduates in terms of discrimination towards women in the workplace. In general, college graduates are more likely to think that women are not discriminated against – this is shown especially in the “Strongly Disagree” category.
Chi-Square Test: Modern Sexism
chisq.test(newdata2$ModernSexism,newdata2$EducationLevel)##
## Pearson's Chi-squared test
##
## data: newdata2$ModernSexism and newdata2$EducationLevel
## X-squared = 45.766, df = 3, p-value = 6.359e-10This chi-squared test shows that there is a statistically significant relationship between education level and their opinion on discrimination towards women in the workforce. The p-value obtained from this test is 6.359e-10 which is less than 0.05, therefore, signifying that there is statistical significance.
Analysis 4
Analyzing relationship between education level (EducationLevel) and feeling towards women on a scale from 1-100 (FeelingAboutWomen).
Table Comparing Means: Feeling Towards Women
newdata3<-newdata2%>%
filter(EducationLevel %in% c("College graduate","High school graduate"))%>%
group_by(EducationLevel)%>%
summarize(Avg_FT_Women = mean(FeelingAboutWomen,na.rm=TRUE))## `summarise()` ungrouping output (override with `.groups` argument)newdata3## # A tibble: 2 x 2
## EducationLevel Avg_FT_Women
## <chr> <dbl>
## 1 College graduate 76.8
## 2 High school graduate 76.9The above table shows the difference in means between college graduates and high school graduates in their average feeling towards women. The means are almost the same, with a very nominal difference of only 0.11059.
Difference in means visualization
ggplot()+
geom_col(data=newdata3,aes(x=EducationLevel,y=Avg_FT_Women,fill=Avg_FT_Women))
The graph further displays how close the means are. The difference in means between college graduates and high school graduates on their feeling towards women is only 0.11059, making the columns appear nearly identical to the human eye. Because the means are so close, I hypothesize that there is no statistically significant difference between high school and college gradutes in their average feeling towards women. We will further analyze this hypothesis below.
Visualization: Histogram comparing population distributions
newdata2%>%
ggplot()+
geom_histogram(aes(x=FeelingAboutWomen))+
facet_wrap(~EducationLevel)## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.## Warning: Removed 172 rows containing non-finite values (stat_bin).
The above graphs depict the population disstribtions for the average feeling towards women between college graduates versus high school graduates. The graphs are relatively the same, except there is a bigger spike at 100 for feeling towards women for high school graduates. Overall, however, these graphs closely resemble one another – especially for population distributions. It should be noted though that neither histogtam for the population distribution follows a normal distribution.
Visualization: Histogram comparing sampling distributions
Creating 2 different datasets to filter out the different education levels being analyzed. These datasets will then be used to create sampling distributions.
Data Set 1 (college graduates)
data1=college_data<-newdata2%>%
filter(EducationLevel=="College graduate")Data Set 2 (high school graduates)
data2=highschool_data<-newdata2%>%
filter(EducationLevel=="High school graduate")Drawing 10,000 random samples of 40 respondents for both data sets shown above then creating a sampling distribution to compare. The sampling distribution shows both datasets on one graph.
data1=college_data<-replicate(10000,
sample(newdata2$FeelingAboutWomen, 40)%>%mean(na.rm=TRUE))%>%
data.frame()%>%
rename("mean"=1)
data2=highschool_data<-replicate(10000,
sample(newdata2$FeelingAboutWomen, 40)%>%mean(na.rm=TRUE)
)%>%
data.frame()%>%
rename("mean"=1)
ggplot()+
geom_histogram(data=college_data,aes(x=mean),fill="red",alpha=0.5)+
geom_histogram(data=highschool_data,aes(x=mean),fill="green",alpha=0.5)## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
From this graph we can see right away that these sampling distributions are far closer to a normal distribution than the population distributions were. The sample distributions also overlap greatly, with only slight differences between the two.
T-Test
Hypotheses
Null Hypothesis: There is no difference in the mean value between the two groups.
Alternative Hypothesis: There is a difference in the mean value between the two groups.
t.test(FeelingAboutWomen~EducationLevel, data=newdata2)##
## Welch Two Sample t-test
##
## data: FeelingAboutWomen by EducationLevel
## t = -0.14775, df = 3105.1, p-value = 0.8826
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -1.578124 1.356953
## sample estimates:
## mean in group College graduate mean in group High school graduate
## 76.78163 76.89222