Data Summarization: Grouped summaries

Nicky Wakim

Grouped summaries

  • Sometimes we want to compare summary statistics for different groups of people in our dataset

  • You can use group_by() to specify the grouping variable(s) before summarizing

  • Then you can use another summary function to calculate the statistics for each group separately

  • We will review the functions that can be used for grouped summaries:

    • summarise() from dplyr
    • skim() from skimr
    • get_summary_stats() from rstatix
    • tbl_summary() from gtsummary
      • Slightly different: does not use group_by()

 

  • For each, we will look at grouped summaries for:
    • Grouped by sex
    • Summarize age_yr, cesd, and income

New summary function: summarise()/summarize()

  • Not necessarily grouped
  • Flexible function that allows you to calculate any summary statistic you want, and name the resulting columns yourself
    • You need to feed it a column and a function to apply to that column

 

  • Really flexible: you can calculate custom summary statistics
  • More work if you want general summary statistics: you have to write out each statistic and name each resulting column yourself

Example: summary statistics on age

  • For each summary statistic, I need to use the designated function and specify the column I want to summarize
hrs_00 |>
  summarise(
    mean_age = mean(age_yr, na.rm = TRUE),
    mean_age_above_80 = mean(age_yr[age_yr > 80], na.rm = TRUE),
    median_age = median(age_yr, na.rm = TRUE),
    sd_age = sd(age_yr, na.rm = TRUE),
    min_age = min(age_yr, na.rm = TRUE),
    max_age = max(age_yr, na.rm = TRUE), 
  )
1
I can create a unique summary statistic without changing the dataset. I have calculated the mean age for people above 80 years old. 
  mean_age mean_age_above_80 median_age   sd_age min_age max_age
1 58.82075          85.73333         55 9.499722      34     100

Group by a variable in summarise()

hrs_00 |>
  group_by(sex) |>
  summarise(
    mean_age = mean(age_yr, na.rm = TRUE),
    sd_age = sd(age_yr, na.rm = TRUE),
    mean_cesd = mean(cesd, na.rm = TRUE),
    sd_cesd = sd(cesd, na.rm = TRUE),
    mean_income = mean(income, na.rm = TRUE),
    sd_income = sd(income, na.rm = TRUE)
  )
1
Use group_by() to specify the grouping variable before summarizing. The summary statistics will be calculated separately for each level of sex.
2
Use summarise() to calculate the summary statistics. Syntax will be the same as non-grouped summaries. 
# A tibble: 2 × 7
  sex    mean_age sd_age mean_cesd sd_cesd mean_income sd_income
  <fct>     <dbl>  <dbl>     <dbl>   <dbl>       <dbl>     <dbl>
1 Female     59.0  10.3       1.92    2.20      82243.   106415.
2 Male       58.6   8.28      1.61    2.02     104748.   152653.

Group by a variable in skim()

hrs_00 |>
  group_by(sex) |>
  select(age_yr, cesd, income) |>
  skim()
1
Use group_by() to specify the grouping variable before summarizing. The summary statistics will be calculated separately for each level of sex.
2
Use select() to specify the variables you want to summarize.
3
Use skim() to calculate the summary statistics. Syntax will be the same as non-grouped summaries.
Data summary
Name select(group_by(hrs_00, s…
Number of rows 2728
Number of columns 4
_______________________
Column type frequency:
numeric 3
________________________
Group variables sex

Variable type: numeric

skim_variable sex n_missing complete_rate mean sd p0 p25 p50 p75 p100 hist
age_yr Female 0 1 58.96 10.31 34 52.00 55.0 65 100 ▁▇▃▁▁
age_yr Male 0 1 58.63 8.28 40 53.00 56.0 64 92 ▁▇▃▁▁
cesd Female 4 1 1.92 2.20 0 0.00 1.0 3 8 ▇▃▁▁▁
cesd Male 3 1 1.61 2.02 0 0.00 1.0 2 8 ▇▂▁▁▁
income Female 0 1 82243.22 106415.42 0 18230.09 47326.0 102785 990200 ▇▁▁▁▁
income Male 0 1 104748.19 152652.90 0 22084.50 58734.8 126650 1800012 ▇▁▁▁▁

Group by a variable in get_summary_stats()

hrs_00 |>
  group_by(sex) |>
  get_summary_stats(
    age_yr, cesd, income, 
    type = "common"
    )
1
Use group_by() to specify the grouping variable before summarizing. The summary statistics will be calculated separately for each level of sex.
2
Use get_summary_stats() to calculate the summary statistics. Syntax will be the same as non-grouped summaries. 
# A tibble: 6 × 11
  sex    variable     n   min    max median    iqr   mean     sd      se      ci
  <fct>  <fct>    <dbl> <dbl>  <dbl>  <dbl>  <dbl>  <dbl>  <dbl>   <dbl>   <dbl>
1 Female age_yr    1568    34 1   e2    55  1.3 e1 5.90e1 1.03e1 2.6 e-1 5.11e-1
2 Female cesd      1564     0 8   e0     1  3   e0 1.92e0 2.20e0 5.6 e-2 1.09e-1
3 Female income    1568     0 9.90e5 47326  8.46e4 8.22e4 1.06e5 2.69e+3 5.27e+3
4 Male   age_yr    1160    40 9.2 e1    56  1.1 e1 5.86e1 8.28e0 2.43e-1 4.77e-1
5 Male   cesd      1157     0 8   e0     1  2   e0 1.61e0 2.02e0 5.9 e-2 1.17e-1
6 Male   income    1160     0 1.80e6 58735. 1.05e5 1.05e5 1.53e5 4.48e+3 8.79e+3

Group by a variable in tbl_summary()

  • Slightly different syntax than the other functions: does not use group_by()
  • by = splits the table into side-by-side columns for each group
hrs_00 |>
  select(age_yr, cesd, income, sex) |>
  tbl_summary(by = sex)
1
Stratify the whole table by sex, creating one column per group
Characteristic Female
N = 1,5681		 Male
N = 1,1601
Age at Interview (Years) 55 (52, 65) 56 (53, 64)
CES-D Depression Score

    0 543 (35%) 450 (39%)
    1 357 (23%) 300 (26%)
    2 211 (13%) 135 (12%)
    3 116 (7.4%) 84 (7.3%)
    4 92 (5.9%) 60 (5.2%)
    5 78 (5.0%) 44 (3.8%)
    6 70 (4.5%) 35 (3.0%)
    7 71 (4.5%) 22 (1.9%)
    8 26 (1.7%) 27 (2.3%)
    Unknown 4 3
Total Household Income 47,326 (18,229, 102,785) 58,735 (22,059, 126,700)
1 Median (Q1, Q3); n (%)

Wrap-up

  • We have covered several functions for summarizing grouped data in R
    • summarise()
    • skim()
    • get_summary_stats()
    • tbl_summary()