• is.na(),any(is.na()), all(is.na()),count(), and functions from naniar like gg_miss_var() and miss_var_summary can help determine if we have NA values
• miss_var_which() can help you drop columns that have any missing values.
• filter() automatically removes NA values
• drop_na() can help you remove NA values
• NA values can change your calculation results
• think about what NA values represent - don’t drop them if you shouldn’t
• replace_na() will replace `NA values with a particular value

• case_when() can recode entire values based on conditions
  • remember case_when() needs TRUE ~ variable to keep values that aren’t specified by conditions, otherwise will be NA
• stringr package has great functions for looking for specific parts of values especially filter() and str_detect() combined
  • also has other useful string manipulation functions like str_replace() and more!
  • separate() can split columns into additional columns
  • unite() can combine columns

📃Cheatsheet

In this module, we will show you how to:

1. Reshape data from wide to long
2. Reshape data from long to wide
3. Merge Data/Joins

Data is stored differently in the tibble.

Here’s a small dataset looking at vaccination rates over three months in Alabama.

Wide: has many columns

# A tibble: 1 × 4
  State   June_vacc_rate May_vacc_rate April_vacc_rate
  <chr>            <dbl>         <dbl>           <dbl>
1 Alabama          0.516         0.514           0.511

Long: column names become data

# A tibble: 3 × 3
  State   name            value
  <chr>   <chr>           <dbl>
1 Alabama June_vacc_rate  0.516
2 Alabama May_vacc_rate   0.514
3 Alabama April_vacc_rate 0.511

Wide: multiple columns per individual, values spread across multiple columns

# A tibble: 2 × 4
  State   June_vacc_rate May_vacc_rate April_vacc_rate
  <chr>            <dbl>         <dbl>           <dbl>
1 Alabama          0.516         0.514           0.511
2 Alaska           0.627         0.626           0.623

Long: multiple rows per observation, a single column contains the values

# A tibble: 6 × 3
  State   name            value
  <chr>   <chr>           <dbl>
1 Alabama June_vacc_rate  0.516
2 Alabama May_vacc_rate   0.514
3 Alabama April_vacc_rate 0.511
4 Alaska  June_vacc_rate  0.627
5 Alaska  May_vacc_rate   0.626
6 Alaska  April_vacc_rate 0.623

https://github.com/gadenbuie/tidyexplain/blob/main/images/tidyr-pivoting.gif

Wide: Easier for humans to read

# A tibble: 2 × 4
  State   June_vacc_rate May_vacc_rate April_vacc_rate
  <chr>            <dbl>         <dbl>           <dbl>
1 Alabama          0.516         0.514           0.511
2 Alaska           0.627         0.626           0.623

Long: Easier for R to make plots & do analysis

# A tibble: 6 × 3
  State   name            value
  <chr>   <chr>           <dbl>
1 Alabama June_vacc_rate  0.516
2 Alabama May_vacc_rate   0.514
3 Alabama April_vacc_rate 0.511
4 Alaska  June_vacc_rate  0.627
5 Alaska  May_vacc_rate   0.626
6 Alaska  April_vacc_rate 0.623

We will be talking about:

• pivot_longer - make multiple columns into variables, (wide to long)
• pivot_wider - make a variable into multiple columns, (long to wide)

The reshape command exists. Its arguments are considered more confusing, so we don’t recommend it.

You might see old functions gather and spread when googling. These are older iterations of pivot_longer and pivot_wider, respectively.

ex_wide

# A tibble: 2 × 4
  State   June_vacc_rate May_vacc_rate April_vacc_rate
  <chr>            <dbl>         <dbl>           <dbl>
1 Alabama          0.516         0.514           0.511
2 Alaska           0.627         0.626           0.623

ex_long <- ex_wide %>% pivot_longer(cols = ends_with("rate"))
ex_long

# A tibble: 6 × 3
  State   name            value
  <chr>   <chr>           <dbl>
1 Alabama June_vacc_rate  0.516
2 Alabama May_vacc_rate   0.514
3 Alabama April_vacc_rate 0.511
4 Alaska  June_vacc_rate  0.627
5 Alaska  May_vacc_rate   0.626
6 Alaska  April_vacc_rate 0.623

What does pivot_longer() do?

A. Summarize data

B. Import data

C. Reshape data

Let’s imagine we want to see what proportion of the population was exposed to different nitrate concentrations. Results should look something like:

# A tibble: 3 × 2
  concentration_cat avg_prop_exposedpop
  <chr>                           <dbl>
1 0-3ug/L                      0.593   
2 10-20ug/L                    0.000678
3 more_than_20ug/L             0.000129

wide_nitrate <- wide_nitrate %>%
  select(!ends_with("exceedances"))
wide_nitrate

# A tibble: 88 × 9
    year quarter pop_on_sampled_PWS `pop_0-3ug/L` `pop_>3-5ug/L` `pop_>5-10ug/L`
   <dbl> <chr>                <dbl>         <dbl>          <dbl>           <dbl>
 1  1999 Q1                  106720         67775              0              32
 2  1999 Q2                   85541         55476              0             212
 3  1999 Q3                  559137        319252         231186             212
 4  1999 Q4                   26995         25969            420               0
 5  2000 Q1                   34793          5904              0              92
 6  2000 Q2                  184521        157396              0              32
 7  2000 Q3                   42081         20407            345               0
 8  2000 Q4                  407219        358828            995             412
 9  2001 Q1                   90054         49552            150               0
10  2001 Q2                   83521         43633           2536              90
# ℹ 78 more rows
# ℹ 3 more variables: `pop_>10-20ug/L` <dbl>, `pop_>20ug/L` <dbl>,
#   `pop_on_PWS_with_non-detect` <dbl>

long_nitrate <- wide_nitrate %>%
  pivot_longer(!c(year, quarter, pop_on_sampled_PWS))
long_nitrate

# A tibble: 528 × 5
    year quarter pop_on_sampled_PWS name                       value
   <dbl> <chr>                <dbl> <chr>                      <dbl>
 1  1999 Q1                  106720 pop_0-3ug/L                67775
 2  1999 Q1                  106720 pop_>3-5ug/L                   0
 3  1999 Q1                  106720 pop_>5-10ug/L                 32
 4  1999 Q1                  106720 pop_>10-20ug/L                 0
 5  1999 Q1                  106720 pop_>20ug/L                    0
 6  1999 Q1                  106720 pop_on_PWS_with_non-detect 38913
 7  1999 Q2                   85541 pop_0-3ug/L                55476
 8  1999 Q2                   85541 pop_>3-5ug/L                   0
 9  1999 Q2                   85541 pop_>5-10ug/L                212
10  1999 Q2                   85541 pop_>10-20ug/L                60
# ℹ 518 more rows

Un-pivoted columns (year, quarter, pop_on_sampled_PWS) are still columns.

long_nitrate

# A tibble: 528 × 5
    year quarter pop_on_sampled_PWS name                       value
   <dbl> <chr>                <dbl> <chr>                      <dbl>
 1  1999 Q1                  106720 pop_0-3ug/L                67775
 2  1999 Q1                  106720 pop_>3-5ug/L                   0
 3  1999 Q1                  106720 pop_>5-10ug/L                 32
 4  1999 Q1                  106720 pop_>10-20ug/L                 0
 5  1999 Q1                  106720 pop_>20ug/L                    0
 6  1999 Q1                  106720 pop_on_PWS_with_non-detect 38913
 7  1999 Q2                   85541 pop_0-3ug/L                55476
 8  1999 Q2                   85541 pop_>3-5ug/L                   0
 9  1999 Q2                   85541 pop_>5-10ug/L                212
10  1999 Q2                   85541 pop_>10-20ug/L                60
# ℹ 518 more rows

Now our data is more tidy, and we can take the averages easily!

long_nitrate %>% 
  group_by(name) %>% 
  summarize("avg_prop_exposedpop" = mean(conc_prop))

# A tibble: 6 × 2
  name                       avg_prop_exposedpop
  <chr>                                    <dbl>
1 pop_0-3ug/L                           0.593   
2 pop_>10-20ug/L                        0.000678
3 pop_>20ug/L                           0.000129
4 pop_>3-5ug/L                          0.182   
5 pop_>5-10ug/L                         0.0189  
6 pop_on_PWS_with_non-detect            0.206   

There are many ways to select the columns we want. Check out https://dplyr.tidyverse.org/reference/dplyr_tidy_select.html to look at more column selection options.

pivot_wider() - spreads row data into columns (tidyr package)

• names_from = the old column whose contents will be spread into multiple new column names.
• values_from = the old column whose contents will fill in the values of those new columns.

{wide_data} <- {long_data} %>%
  pivot_wider(names_from = {Old column name: contains new column names},
              values_from = {Old column name: contains new cell values})

We can use pivot_wider to convert long data to wide format. Let’s try it with the vaccine data from earlier.

ex_long

# A tibble: 6 × 3
  State   Month            Rate
  <chr>   <chr>           <dbl>
1 Alabama June_vacc_rate  0.516
2 Alabama May_vacc_rate   0.514
3 Alabama April_vacc_rate 0.511
4 Alaska  June_vacc_rate  0.627
5 Alaska  May_vacc_rate   0.626
6 Alaska  April_vacc_rate 0.623

We can use pivot_wider to convert long data to wide format. Let’s try it with the vaccine data from earlier.

ex_wide2 <- ex_long %>% pivot_wider(names_from = "Month", 
                                       values_from = "Rate") 
ex_wide2

# A tibble: 2 × 4
  State   June_vacc_rate May_vacc_rate April_vacc_rate
  <chr>            <dbl>         <dbl>           <dbl>
1 Alabama          0.516         0.514           0.511
2 Alaska           0.627         0.626           0.623

Drop some columns we don’t need.

long_nitrate <- long_nitrate %>%
  select(!c(pop_on_sampled_PWS, value))
long_nitrate

# A tibble: 528 × 4
    year quarter name                       conc_prop
   <dbl> <chr>   <chr>                          <dbl>
 1  1999 Q1      pop_0-3ug/L                 0.635   
 2  1999 Q1      pop_>3-5ug/L                0       
 3  1999 Q1      pop_>5-10ug/L               0.000300
 4  1999 Q1      pop_>10-20ug/L              0       
 5  1999 Q1      pop_>20ug/L                 0       
 6  1999 Q1      pop_on_PWS_with_non-detect  0.365   
 7  1999 Q2      pop_0-3ug/L                 0.649   
 8  1999 Q2      pop_>3-5ug/L                0       
 9  1999 Q2      pop_>5-10ug/L               0.00248 
10  1999 Q2      pop_>10-20ug/L              0.000701
# ℹ 518 more rows

Pivot the data!

wide_nitrate <- long_nitrate %>%
  pivot_wider(names_from = "quarter", values_from = "conc_prop")
wide_nitrate

# A tibble: 132 × 6
    year name                             Q1       Q2       Q3      Q4
   <dbl> <chr>                         <dbl>    <dbl>    <dbl>   <dbl>
 1  1999 pop_0-3ug/L                0.635    0.649    0.571    0.962  
 2  1999 pop_>3-5ug/L               0        0        0.413    0.0156 
 3  1999 pop_>5-10ug/L              0.000300 0.00248  0.000379 0      
 4  1999 pop_>10-20ug/L             0        0.000701 0        0      
 5  1999 pop_>20ug/L                0        0        0        0      
 6  1999 pop_on_PWS_with_non-detect 0.365    0.348    0.0152   0.0224 
 7  2000 pop_0-3ug/L                0.170    0.853    0.485    0.881  
 8  2000 pop_>3-5ug/L               0        0        0.00820  0.00244
 9  2000 pop_>5-10ug/L              0.00264  0.000173 0        0.00101
10  2000 pop_>10-20ug/L             0        0        0        0      
# ℹ 122 more rows

• tidyr package helps us convert between wide and long data
• pivot_longer() goes from wide -> long
  • Specify columns you want to pivot
  • Specify names_to = and values_to = for custom naming
• pivot_wider() goes from long -> wide
  • Specify names_from = and values_from =

🏠 Class Website

💻 Lab

“Combining datasets”

• Merging/joining data sets together - usually on key variables, usually “id”
• ?join - see different types of joining for dplyr
• inner_join(x, y) - only rows that match for x and y are kept
• full_join(x, y) - all rows of x and y are kept
• left_join(x, y) - all rows of x are kept even if not merged with y
• right_join(x, y) - all rows of y are kept even if not merged with x
• anti_join(x, y) - all rows from x not in y keeping just columns from x.

data_As <- read_csv(
  file = "https://daseh.org/data/data_As_1.csv")
data_cold <- read_csv(
  file = "https://daseh.org/data/data_cold_1.csv")

data_As

# A tibble: 2 × 3
  State   June_vacc_rate May_vacc_rate
  <chr>            <dbl>         <dbl>
1 Alabama          0.516         0.514
2 Alaska           0.627         0.626

data_cold

# A tibble: 2 × 2
  State  April_vacc_rate
  <chr>            <dbl>
1 Maine            0.795
2 Alaska           0.623

https://github.com/gadenbuie/tidyexplain/blob/main/images/inner-join.gif

ij <- inner_join(data_As, data_cold)

Joining with `by = join_by(State)`

ij

# A tibble: 1 × 4
  State  June_vacc_rate May_vacc_rate April_vacc_rate
  <chr>           <dbl>         <dbl>           <dbl>
1 Alaska          0.627         0.626           0.623

https://raw.githubusercontent.com/gadenbuie/tidyexplain/main/images/left-join.gif

“Everything to the left of the comma”

lj <- left_join(data_As, data_cold)

Joining with `by = join_by(State)`

lj

# A tibble: 2 × 4
  State   June_vacc_rate May_vacc_rate April_vacc_rate
  <chr>            <dbl>         <dbl>           <dbl>
1 Alabama          0.516         0.514          NA    
2 Alaska           0.627         0.626           0.623

# install.packages("tidylog")
library(tidylog)
left_join(data_As, data_cold)

Joining with `by = join_by(State)`
left_join: added one column (April_vacc_rate)
> rows only in data_As 1
> rows only in data_cold (1)
> matched rows 1
> ===
> rows total 2

# A tibble: 2 × 4
  State   June_vacc_rate May_vacc_rate April_vacc_rate
  <chr>            <dbl>         <dbl>           <dbl>
1 Alabama          0.516         0.514          NA    
2 Alaska           0.627         0.626           0.623

https://raw.githubusercontent.com/gadenbuie/tidyexplain/main/images/right-join.gif

“Everything to the right of the comma”

rj <- right_join(data_As, data_cold)

Joining with `by = join_by(State)`
right_join: added one column (April_vacc_rate)
> rows only in data_As (1)
> rows only in data_cold 1
> matched rows 1
> ===
> rows total 2

rj

# A tibble: 2 × 4
  State  June_vacc_rate May_vacc_rate April_vacc_rate
  <chr>           <dbl>         <dbl>           <dbl>
1 Alaska          0.627         0.626           0.623
2 Maine          NA            NA               0.795

lj2 <- left_join(data_cold, data_As)

Joining with `by = join_by(State)`
left_join: added 2 columns (June_vacc_rate, May_vacc_rate)
> rows only in data_cold 1
> rows only in data_As (1)
> matched rows 1
> ===
> rows total 2

lj2

# A tibble: 2 × 4
  State  April_vacc_rate June_vacc_rate May_vacc_rate
  <chr>            <dbl>          <dbl>         <dbl>
1 Maine            0.795         NA            NA    
2 Alaska           0.623          0.627         0.626

https://raw.githubusercontent.com/gadenbuie/tidyexplain/main/images/full-join.gif