Data Summarization: Tables

Author

Nicky Wakim

Warning: Missing `trust` will be set to FALSE by default for RDS in 2.0.0.

What are summary tables?

Summary tables are a way to present data in a structured format
The use is a little vague here…
- It could mean:
  - A table for categorical data (like a frequency table)
  - A presentable table for categorical and numeric data

We will cover two functions that can help us create summary tables for categorical and numeric data:
- tabyl() from the janitor package
- tbl_summary() from the gtsummary package

What does `tabyl()` do?

tabyl() is a simple function that creates a frequency table for a single categorical variable
It returns a data frame with three columns:
- n: the count in each category
- percent: that count as a proportion of all rows (including missing values)
- valid_percent: that count as a proportion of only the non-missing rows

Pretty simple to use and can be used to create two-by-two tables
Need to feed it one variable at a time

What does `tabyl()` output look like?

hrs_00 |>
  tabyl(degree)

                              degree   n    percent valid_percent
                 High school diploma 512 0.18768328    0.22674934
                     Master's degree 487 0.17851906    0.21567759
 Professional degree (Ph.D./M.D./JD) 194 0.07111437    0.08591674
                                None 384 0.14076246    0.17006200
                  Associate's degree 411 0.15065982    0.18201949
                                 GED 118 0.04325513    0.05225864
                   Bachelor's degree 152 0.05571848    0.06731621
                                <NA> 470 0.17228739            NA

Cross-tabulations with `tabyl()`

You can also cross-tabulate two variables by adding a second column name
- This will give you a table of counts for each combination of the two variables

hrs_00 |>
  tabyl(degree, sex)

                              degree Female Male
                 High school diploma    281  231
                     Master's degree    286  201
 Professional degree (Ph.D./M.D./JD)    115   79
                                None    238  146
                  Associate's degree    225  186
                                 GED     60   58
                   Bachelor's degree     88   64
                                <NA>    275  195

Dressing up a `tabyl()` with `adorn_*()`

adorn_*() functions let you format that table for presentation
These are designed to be piped together, one after another

Function	What it does
`adorn_totals()`	Adds a total row and/or column (`where = "row"`, `"col"`, or `"both"`)
`adorn_percentages()`	Converts counts to proportions (`denominator = "row"`, `"col"`, or `"all"`)
`adorn_pct_formatting()`	Converts proportions to nicely formatted percents (control decimals with `digits =`)
`adorn_ns()`	Adds the raw counts back onto a table of percents, e.g. `"45.2% (120)"`

Example: Formatting a frequency table

We can add totals, convert to percentages, format as percents, and add counts back to a frequency table
Order matters!
- You generally want: tabulate → add totals → convert to percentages → format as percents → add counts back

hrs_00 |>
  tabyl(degree) |>
1  adorn_totals(where = "row") |>
2  adorn_pct_formatting(digits = 2)

1: Add a row for totals to the table. A little confusing because it totals each column, but “row” refers to the position of the total row, not what is being totaled
2: Format the percentages to 2 decimal places

                              degree    n percent valid_percent
                 High school diploma  512  18.77%        22.67%
                     Master's degree  487  17.85%        21.57%
 Professional degree (Ph.D./M.D./JD)  194   7.11%         8.59%
                                None  384  14.08%        17.01%
                  Associate's degree  411  15.07%        18.20%
                                 GED  118   4.33%         5.23%
                   Bachelor's degree  152   5.57%         6.73%
                                <NA>  470  17.23%             -
                               Total 2728 100.00%       100.00%

Example: Formatting a cross-tabulation

We can also add totals, convert to percentages, format as percents, and add counts back to a cross-tabulation

hrs_00 |>
1  tabyl(degree, sex) |>
2  adorn_totals(where = "row") |>
3  adorn_percentages(denominator = "col") |>
4  adorn_pct_formatting(digits = 1) |>
5  adorn_ns(position = "front")

1: Create a cross-tabulation of degree by sex
2: Add a row for totals to the table. A little confusing because it totals each column.
3: Convert the counts to percentages, using the column totals (the added row of totals) as the denominator.
4: Format the percentages to 1 decimal place
5: Add the raw counts back to the table, in front of the percentages.

                              degree         Female           Male
                 High school diploma   281  (17.9%)   231  (19.9%)
                     Master's degree   286  (18.2%)   201  (17.3%)
 Professional degree (Ph.D./M.D./JD)   115   (7.3%)    79   (6.8%)
                                None   238  (15.2%)   146  (12.6%)
                  Associate's degree   225  (14.3%)   186  (16.0%)
                                 GED    60   (3.8%)    58   (5.0%)
                   Bachelor's degree    88   (5.6%)    64   (5.5%)
                                <NA>   275  (17.5%)   195  (16.8%)
                               Total 1,568 (100.0%) 1,160 (100.0%)

What does `tbl_summary()` do?

tbl_summary() builds a publication-ready summary table for all the columns you give it in a single command
It automatically picks sensible statistics based on each variable’s type:
- Numeric variables: median (interquartile range)
- Categorical variables: count (percent)
- Missing values are shown as their own “Unknown” row
We need to use select() to decide what goes into the table

Great for publications!
You can customize the statistics shown, the labels, and even stratify the table by a grouping variable
Some of the customizations are syntax heavy

What does `tbl_summary()` output look like?

hrs_00 |>
1  select(-id, -hhid, -pn) |>
  tbl_summary()

1: Only removing this because it would add thousands of rows to this table! See what happens when you don’t remove these variables!

Characteristic	N = 2,728¹
Ever Had Diabetes	735 (27%)
Highest Degree Earned
High school diploma	512 (23%)
Master's degree	487 (22%)
Professional degree (Ph.D./M.D./JD)	194 (8.6%)
None	384 (17%)
Associate's degree	411 (18%)
GED	118 (5.2%)
Bachelor's degree	152 (6.7%)
Unknown	470
Respondent Birth Year	1,967 (1,958, 1,970)
Respondent Birth Month	7.0 (3.5, 10.0)
Respondent Birth Date	2,602 (-489, 3,757)
Proxy Interview Status
Respondent	2,721 (100%)
Proxy	7 (0.3%)
Couple Household Status
Not a couple HH	1,093 (40%)
Couple HH	1,635 (60%)
Sex
Female	1,568 (57%)
Male	1,160 (43%)
Age at Interview (Months)	671 (635, 778)
Age at Interview (Years)	55 (52, 64)
Years of Education	13.0 (12.0, 16.0)
Unknown	507
Race/Ethnicity
White/Caucasian	1,000 (45%)
Black/African American	884 (39%)
Other	358 (16%)
Unknown	486
Current Smoker	459 (17%)
Unknown	2
Ever Smoked	1,243 (46%)
Unknown	1
Current Drinker	1,874 (69%)
Self-Reported Height	1.68 (1.63, 1.77)
Unknown	55
Self-Reported Health Status
Fair	743 (27%)
Good	870 (32%)
Very Good	647 (24%)
Excellent	277 (10%)
Poor	187 (6.9%)
Unknown	4
Vigorous Physical Activity Frequency
Never	1,443 (53%)
>1 per week	586 (22%)
1 per week	270 (9.9%)
1-3 per week	275 (10%)
Every day	145 (5.3%)
Unknown	9
Ever Had High Blood Pressure	1,433 (53%)
Unknown	8
Ever Had Cancer	271 (9.9%)
Ever Had Lung Disease	160 (5.9%)
Ever Had Heart Problems	373 (14%)
Ever Had Stroke	193 (7.1%)
Ever Had Psychiatric Problems	540 (20%)
Ever Had Sleep Disorder	534 (20%)
Ever Had Arthritis	1,006 (37%)
Total Number of Chronic Conditions
0	635 (23%)
1	675 (25%)
2	651 (24%)
3	443 (16%)
4	227 (8.3%)
5	71 (2.6%)
6	23 (0.8%)
7	3 (0.1%)
CES-D Depression Score
0	993 (36%)
1	657 (24%)
2	346 (13%)
3	200 (7.4%)
4	152 (5.6%)
5	122 (4.5%)
6	105 (3.9%)
7	93 (3.4%)
8	53 (1.9%)
Unknown	7
Total Household Income	51,462 (19,852, 116,500)
¹ n (%); Median (Q1, Q3)

Customizing `tbl_summary()`

statistic = lets you change which statistics are shown, using { } to insert stat names like {mean}, {sd}, {median}, {n}, {p} (percent)
label = lets you rename variables to something more readable

hrs_00 |>
  select(age_yr, sex, diab) |>
  tbl_summary(
1    statistic = age_yr ~ "{mean} ({sd})",
2    label = age_yr ~ "Age (years)"
  )

1: Show mean (sd) for age_yr instead of the default median (IQR)
2: Give age_yr a nicer display name

Characteristic	N = 2,728¹
Age (years)	59 (9)
Sex
Female	1,568 (57%)
Male	1,160 (43%)
Ever Had Diabetes	735 (27%)
¹ Mean (SD); n (%)

Presentation-ready tables with `gt()`

gt() turns any data frame or table (including a tabyl() output) into a polished, presentation-ready table
It’s a general-purpose “make this table look nice” tool
You can further customize titles, column labels, and colors using additional gt functions (like tab_header() or fmt_percent())

Original table from tabyl()

hrs_00 |>
  tabyl(degree)

                              degree   n    percent valid_percent
                 High school diploma 512 0.18768328    0.22674934
                     Master's degree 487 0.17851906    0.21567759
 Professional degree (Ph.D./M.D./JD) 194 0.07111437    0.08591674
                                None 384 0.14076246    0.17006200
                  Associate's degree 411 0.15065982    0.18201949
                                 GED 118 0.04325513    0.05225864
                   Bachelor's degree 152 0.05571848    0.06731621
                                <NA> 470 0.17228739            NA

Using gt() to make it look nice

hrs_00 |>
  tabyl(degree) |>
  gt() |>
  tab_header(title = "Highest Degree Earned")

degree	n	percent	valid_percent
Highest Degree Earned
High school diploma	512	0.18768328	0.22674934
Master's degree	487	0.17851906	0.21567759
Professional degree (Ph.D./M.D./JD)	194	0.07111437	0.08591674
None	384	0.14076246	0.17006200
Associate's degree	411	0.15065982	0.18201949
GED	118	0.04325513	0.05225864
Bachelor's degree	152	0.05571848	0.06731621
NA	470	0.17228739	NA

We can change the font size for `tbl_summary()`

hrs_00 |>
  select(age_yr, sex, diab) |>
  tbl_summary(
    statistic = age_yr ~ "{mean} ({sd})",          
    label = age_yr ~ "Age (years)"                 
  ) |>
1  as_gt() |>
2  tab_options(table.font.size = px(50))

1: Convert the tbl_summary() output to a gt table so we can customize it. Need to use as_gt() instead of gt() because tbl_summary() already creates a table.
2: Change the font size to 50 pixels. Be careful: font size is relative to your document. 50px is probably too large for your standard html’s.

Characteristic	N = 2,728¹
Age (years)	59 (9)
Sex
Female	1,568 (57%)
Male	1,160 (43%)
Ever Had Diabetes	735 (27%)
¹ Mean (SD); n (%)

Wrap-up

We covered a few functions for summarizing data in R:
- tabyl() from the janitor package
  - For categorical data
  - Can be used for one-way or two-way tables
- tbl_summary() from the gtsummary package
  - For categorical and numeric data
  - Good for publications!
  - Can be customized to show different statistics, labels, and stratifications

Resources

Tutorial on tbl_summary() from the gtsummary package
- Includes more options than covered here!
More on the gt package
- Includes more options than covered here!

What are summary tables?

What does tabyl() do?

What does tabyl() output look like?

Cross-tabulations with tabyl()

Dressing up a tabyl() with adorn_*()

Example: Formatting a frequency table

Example: Formatting a cross-tabulation

What does tbl_summary() do?

What does tbl_summary() output look like?

Customizing tbl_summary()

Presentation-ready tables with gt()

We can change the font size for tbl_summary()

Wrap-up

Resources

What does `tabyl()` do?

What does `tabyl()` output look like?

Cross-tabulations with `tabyl()`

Dressing up a `tabyl()` with `adorn_*()`

What does `tbl_summary()` do?

What does `tbl_summary()` output look like?

Customizing `tbl_summary()`

Presentation-ready tables with `gt()`

We can change the font size for `tbl_summary()`