Sono Shah: Using dplyr with the CCES

Sono Shah

Data exploration

There will be many instances in which you want to get an idea of what the people in your dataset look like. For instance, let’s take a look at the gender and levels of education in the CCES dataset. There are two variables that we will work with:

gender = Gender
educ = Educational Attainment

Count of gender

cces_data %>% # I want to look at the cces_data
  count(gender) # Show me a count of gender

  gender     n
1   Male 29531
2 Female 35069

Another way

Here’s another way to do it:

cces_data %>% # I want to look at the cces_data
  group_by(gender) %>% # Group the data by gender 
  summarise(n = n()) # Give me a count

# A tibble: 2 x 2
  gender     n
* <ord>  <int>
1 Male   29531
2 Female 35069

Education?

cces_data %>% # I want to look at the cces_data
  count(educ) # Show me a count of education

                  educ     n
1                No HS  1971
2 High school graduate 16381
3         Some college 15685
4               2-year  7169
5               4-year 14884
6            Post-grad  8510

Education another way

Here’s another way:

cces_data %>% # I want to look at the cces_data
  group_by(educ) %>% # Group the data by education 
  summarise(n = n()) # Give me a count

# A tibble: 6 x 2
  educ                     n
* <ord>                <int>
1 No HS                 1971
2 High school graduate 16381
3 Some college         15685
4 2-year                7169
5 4-year               14884
6 Post-grad             8510

What about both?

What if I want to see gender by education level? This will help us get an idea of the levels of education are consistent, or similar accross gender.

Gender by Education

cces_data %>% # I want to look at the cces_data
  count(gender, educ) # Show me a count of gender BY educational level

   gender                 educ    n
1    Male                No HS  756
2    Male High school graduate 6642
3    Male         Some college 7050
4    Male               2-year 2995
5    Male               4-year 7431
6    Male            Post-grad 4657
7  Female                No HS 1215
8  Female High school graduate 9739
9  Female         Some college 8635
10 Female               2-year 4174
11 Female               4-year 7453
12 Female            Post-grad 3853

Gender by Education another way

cces_data %>% # I want to look at the cces_data
  group_by(gender, educ) %>% # Group the data by gender then education
  summarise(n = n()) # Show me a count of gender BY educational level

# A tibble: 12 x 3
# Groups:   gender [2]
   gender educ                     n
   <ord>  <ord>                <int>
 1 Male   No HS                  756
 2 Male   High school graduate  6642
 3 Male   Some college          7050
 4 Male   2-year                2995
 5 Male   4-year                7431
 6 Male   Post-grad             4657
 7 Female No HS                 1215
 8 Female High school graduate  9739
 9 Female Some college          8635
10 Female 2-year                4174
11 Female 4-year                7453
12 Female Post-grad             3853

Proportions and Percents

This information is ok, but limited. Knowing that there are 756 males in the sample that have No HS diploma, is not that useful. What would be better is to know what proportion or percentage of males in the sample have no HS diploma. This is useful because the sample is nationally representative, meaning it is like a snapshot of the US population. So by using a proportion, we can get an idea of what the real US population looks like on this particular measure.

Gender by Education Proportions

Look we can see that .22 or 22% of males in the CCES dataset have a HS diploma

cces_data %>% # I want to look at the cces_data
  count(gender, educ) %>% # Show me a count of gender by educational level
  group_by(gender) %>% # I want to group by Gender, since I'm interested in the Proportion (percentage) of MALES
  mutate(proportion = n /sum(n)) # Create a variable that takes the raw count then divides it by the total for each gender.

# A tibble: 12 x 4
# Groups:   gender [2]
   gender educ                     n proportion
   <ord>  <ord>                <int>      <dbl>
 1 Male   No HS                  756     0.0256
 2 Male   High school graduate  6642     0.225 
 3 Male   Some college          7050     0.239 
 4 Male   2-year                2995     0.101 
 5 Male   4-year                7431     0.252 
 6 Male   Post-grad             4657     0.158 
 7 Female No HS                 1215     0.0346
 8 Female High school graduate  9739     0.278 
 9 Female Some college          8635     0.246 
10 Female 2-year                4174     0.119 
11 Female 4-year                7453     0.213 
12 Female Post-grad             3853     0.110

Gender by Education Proportions another way

cces_data %>% # I want to look at the cces_data 
  group_by(gender,educ) %>% # I want to group by Gender, since I'm interested in the Proportion (percentage) of MALES
  summarise( n = n()) %>% # Give me a count
  mutate(proportion = n /sum(n)) # Create a variable that takes the raw count then divides it by the total for each gender.

# A tibble: 12 x 4
# Groups:   gender [2]
   gender educ                     n proportion
   <ord>  <ord>                <int>      <dbl>
 1 Male   No HS                  756     0.0256
 2 Male   High school graduate  6642     0.225 
 3 Male   Some college          7050     0.239 
 4 Male   2-year                2995     0.101 
 5 Male   4-year                7431     0.252 
 6 Male   Post-grad             4657     0.158 
 7 Female No HS                 1215     0.0346
 8 Female High school graduate  9739     0.278 
 9 Female Some college          8635     0.246 
10 Female 2-year                4174     0.119 
11 Female 4-year                7453     0.213 
12 Female Post-grad             3853     0.110

What about the proportions/percentages?

Somtimes you don’t need to know the full breakdown of education and you really just want to know for each gender, what level of education do most people in your data have? All we need to to get the level of education for each gender that has the largest proportion.

Largest Proportion

Sweet! This tells you that most of the males in the dataset have a 4-year degree, while most of the females in the dataset are high school graduates.

cces_data %>% # I want to look at the CCES dataset 
  group_by(gender, educ) %>% # Group by gender and education level
  summarise(n = n()) %>% # give me a count
  mutate(proportion = n /sum(n)) %>% # give me the proportion
  top_n(1) # give me the top proportion for each gender

# A tibble: 2 x 4
# Groups:   gender [2]
  gender educ                     n proportion
  <ord>  <ord>                <int>      <dbl>
1 Male   4-year                7431      0.252
2 Female High school graduate  9739      0.278

Largest Proportion another way

cces_data %>% # I want to look at the CCES dataset 
  count(gender, educ) %>% # Group by gender and education level
  group_by(gender) %>% # give me a count
  mutate(proportion = n /sum(n)) %>% # give me the proportion
  top_n(1) # give me the top proportion for each gender

# A tibble: 2 x 4
# Groups:   gender [2]
  gender educ                     n proportion
  <ord>  <ord>                <int>      <dbl>
1 Male   4-year                7431      0.252
2 Female High school graduate  9739      0.278

Smallest Proportion

We can also do the smallest proportion too! Looks like for both males and females, those with NO HS diploma make up the smallest proportion of each group.

cces_data %>% # I want to look at the CCES dataset 
  count(gender, educ) %>% # Group by gender and education level
  group_by(gender) %>% # give me a count
  mutate(proportion = n /sum(n)) %>% # give me the proportion
  top_n(-1) # give me the smallest proportion for each gender

# A tibble: 2 x 4
# Groups:   gender [2]
  gender educ      n proportion
  <ord>  <ord> <int>      <dbl>
1 Male   No HS   756     0.0256
2 Female No HS  1215     0.0346

Using dplyr with the CCES