Zheer Kejlberg Al-Mashhadi 2023-11-19
In this document, the functions in the package are presented with examples. First, the round_5_gtsummary() function is demonstrated, followed by the add_SMD() function
install.packages("devtools")
devtools::install_github("zheer-kejlberg/Z.gtsummary.addons")
library(Z.gtsummary.addons)For one SMD per variable, the location argument does not need to be specified (as it defaults to “label”).
trial %>%
tbl_summary(by = grade, include = c(age, stage)) %>%
add_SMD()| Characteristic | I, N = 681 | II, N = 681 | III, N = 641 | SMD: I vs. II | SMD: I vs. III | SMD: II vs. III |
|---|---|---|---|---|---|---|
| Age | 47 (37, 56) | 49 (37, 57) | 47 (38, 58) | -0.10 | -0.13 | -0.04 |
| Unknown | 2 | 6 | 3 | |||
| T Stage | 0.29 | 0.19 | 0.31 | |||
| T1 | 17 (25%) | 23 (34%) | 13 (20%) | |||
| T2 | 18 (26%) | 17 (25%) | 19 (30%) | |||
| T3 | 18 (26%) | 11 (16%) | 14 (22%) | |||
| T4 | 15 (22%) | 17 (25%) | 18 (28%) | |||
| 1 Median (IQR); n (%) | ||||||
For one SMD per level of every categorical variable, you must specify location = “level”
trial %>%
tbl_summary(by = grade, include = c(age, stage)) %>%
add_SMD(location = "level")| Characteristic | I, N = 681 | II, N = 681 | III, N = 641 | SMD: I vs. II | SMD: I vs. III | SMD: II vs. III |
|---|---|---|---|---|---|---|
| Age | 47 (37, 56) | 49 (37, 57) | 47 (38, 58) | |||
| Unknown | 2 | 6 | 3 | |||
| T Stage | ||||||
| T1 | 17 (25%) | 23 (34%) | 13 (20%) | -0.19 | 0.11 | 0.31 |
| T2 | 18 (26%) | 17 (25%) | 19 (30%) | 0.03 | -0.07 | -0.11 |
| T3 | 18 (26%) | 11 (16%) | 14 (22%) | 0.25 | 0.11 | -0.15 |
| T4 | 15 (22%) | 17 (25%) | 18 (28%) | -0.07 | -0.14 | -0.07 |
| 1 Median (IQR); n (%) | ||||||
There’s also the option to set location = “both” to get both kinds of SMDs simultaneously.
trial %>%
tbl_summary(by = grade, include = c(age, stage)) %>%
add_SMD(location = "both")| Characteristic | I, N = 681 | II, N = 681 | III, N = 641 | SMD: I vs. II | SMD: I vs. III | SMD: II vs. III |
|---|---|---|---|---|---|---|
| Age | 47 (37, 56) | 49 (37, 57) | 47 (38, 58) | -0.10 | -0.13 | -0.04 |
| Unknown | 2 | 6 | 3 | |||
| T Stage | 0.29 | 0.19 | 0.31 | |||
| T1 | 17 (25%) | 23 (34%) | 13 (20%) | -0.19 | 0.11 | 0.31 |
| T2 | 18 (26%) | 17 (25%) | 19 (30%) | 0.03 | -0.07 | -0.11 |
| T3 | 18 (26%) | 11 (16%) | 14 (22%) | 0.25 | 0.11 | -0.15 |
| T4 | 15 (22%) | 17 (25%) | 18 (28%) | -0.07 | -0.14 | -0.07 |
| 1 Median (IQR); n (%) | ||||||
To get confidence intervals, add ci = TRUE. With the decimals argument, you can adjust the number of significant digits displayed.
trial %>%
tbl_summary(by = grade, include = c(age, stage)) %>%
add_SMD(location = "level", ci = TRUE, decimals = 3)| Characteristic | I, N = 681 | II, N = 681 | III, N = 641 | SMD: I vs. II | SMD: I vs. III | SMD: II vs. III |
|---|---|---|---|---|---|---|
| Age | 47 (37, 56) | 49 (37, 57) | 47 (38, 58) | |||
| Unknown | 2 | 6 | 3 | |||
| T Stage | ||||||
| T1 | 17 (25%) | 23 (34%) | 13 (20%) | -0.195 (-0.531, 0.142) | 0.112 (-0.229, 0.454) | 0.308 (-0.036, 0.651) |
| T2 | 18 (26%) | 17 (25%) | 19 (30%) | 0.034 (-0.303, 0.370) | -0.072 (-0.413, 0.270) | -0.105 (-0.447, 0.236) |
| T3 | 18 (26%) | 11 (16%) | 14 (22%) | 0.253 (-0.084, 0.591) | 0.107 (-0.234, 0.449) | -0.146 (-0.487, 0.196) |
| T4 | 15 (22%) | 17 (25%) | 18 (28%) | -0.069 (-0.406, 0.267) | -0.140 (-0.482, 0.202) | -0.071 (-0.412, 0.271) |
| 1 Median (IQR); n (%) | ||||||
To change the formatting of the confidence intervals, use the ci_bracket and ci_sep arguments:
trial %>%
tbl_summary(by = grade, include = c(age, stage)) %>%
add_SMD(location = "level", ci = TRUE, ci_bracket="[]", ci_sep=";")| Characteristic | I, N = 681 | II, N = 681 | III, N = 641 | SMD: I vs. II | SMD: I vs. III | SMD: II vs. III |
|---|---|---|---|---|---|---|
| Age | 47 (37, 56) | 49 (37, 57) | 47 (38, 58) | |||
| Unknown | 2 | 6 | 3 | |||
| T Stage | ||||||
| T1 | 17 (25%) | 23 (34%) | 13 (20%) | -0.19 [-0.53;0.14] | 0.11 [-0.23;0.45] | 0.31 [-0.04;0.65] |
| T2 | 18 (26%) | 17 (25%) | 19 (30%) | 0.03 [-0.30;0.37] | -0.07 [-0.41;0.27] | -0.11 [-0.45;0.24] |
| T3 | 18 (26%) | 11 (16%) | 14 (22%) | 0.25 [-0.08;0.59] | 0.11 [-0.23;0.45] | -0.15 [-0.49;0.20] |
| T4 | 15 (22%) | 17 (25%) | 18 (28%) | -0.07 [-0.41;0.27] | -0.14 [-0.48;0.20] | -0.07 [-0.41;0.27] |
| 1 Median (IQR); n (%) | ||||||
In this example, we use weights from WeightIt package. The survey package delivers the necessary svydesign object.
library(WeightIt) # To calculate weights
library(survey) # To create a surveydesign object (a "weighted" dataset)Application of the add_SMD() function is identical to the non-weighted case, but it is applied to a tbl_svysummary object instead of a tbl_summary object.
trial %>% mutate(
w = weightit(grade ~ age + stage + trt, data = ., focal="I")$weights) %>% # create ATT weights
survey::svydesign(~1, data = ., weights = ~w) %>% # create the svydesign object
tbl_svysummary(by = grade, include = c(age, stage)) %>%
add_SMD(ref_group = TRUE)| Characteristic | I, N = 681 | II, N = 681 | III, N = 681 | SMD: I vs. II | SMD: I vs. III |
|---|---|---|---|---|---|
| Age | 47 (37, 56) | 46 (34, 56) | 45 (38, 54) | -0.01 | -0.01 |
| Unknown | 2 | 2 | 2 | ||
| T Stage | 0.08 | 0.02 | |||
| T1 | 17 (25%) | 19 (27%) | 17 (25%) | ||
| T2 | 18 (26%) | 19 (27%) | 17 (26%) | ||
| T3 | 18 (26%) | 17 (26%) | 18 (27%) | ||
| T4 | 15 (22%) | 13 (19%) | 15 (23%) | ||
| 1 Median (IQR); n (%) | |||||
*Notice, comparisons are only made here between group I and all other groups due to the use of ref_group = TRUE
The function can be run on both tbl_summary and tbl_svysummary objects.
For comparison, here’s a table without rounding:
trial %>%
tbl_summary(by = grade, include = c(trt, age, stage))| Characteristic | I, N = 681 | II, N = 681 | III, N = 641 |
|---|---|---|---|
| Chemotherapy Treatment | |||
| Drug A | 35 (51%) | 32 (47%) | 31 (48%) |
| Drug B | 33 (49%) | 36 (53%) | 33 (52%) |
| Age | 47 (37, 56) | 49 (37, 57) | 47 (38, 58) |
| Unknown | 2 | 6 | 3 |
| T Stage | |||
| T1 | 17 (25%) | 23 (34%) | 13 (20%) |
| T2 | 18 (26%) | 17 (25%) | 19 (30%) |
| T3 | 18 (26%) | 11 (16%) | 14 (22%) |
| T4 | 15 (22%) | 17 (25%) | 18 (28%) |
| 1 n (%); Median (IQR) | |||
Now, the same table but with all counts rounded to nearest 5 (and all proportions adjusted accordingly):
trial %>%
tbl_summary(by = grade, include = c(trt, age, stage)) %>%
round_5_gtsummary()| Characteristic | I, N = 701 | II, N = 701 | III, N = 651 |
|---|---|---|---|
| Chemotherapy Treatment | |||
| Drug A | 35 (50%) | 30 (42.9%) | 30 (46.2%) |
| Drug B | 35 (50%) | 35 (50%) | 35 (53.8%) |
| Age | 47 (37, 56) | 49 (37, 57) | 47 (38, 58) |
| Unknown | <5 (<7.1%) | 5 (7.1%) | <5 (<7.7%) |
| T Stage | |||
| T1 | 15 (21.4%) | 25 (35.7%) | 15 (23.1%) |
| T2 | 20 (28.6%) | 15 (21.4%) | 20 (30.8%) |
| T3 | 20 (28.6%) | 10 (14.3%) | 15 (23.1%) |
| T4 | 15 (21.4%) | 15 (21.4%) | 20 (30.8%) |
| 1 n (%); Median (IQR) | |||
Finally, the function can also be applied to weighted data:
library(WeightIt) # for calculating weights
weighted_trial <- trial %>%
mutate(w = weightit(grade ~ trt + age + stage, estimand = "ATT", focal = "I")$weights) %>%
survey::svydesign(~1, data = ., weights = ~w)
weighted_trial %>%
tbl_svysummary(by = grade, include = c(trt, age, stage)) %>%
round_5_gtsummary()| Characteristic | I, N = 701 | II, N = 701 | III, N = 701 |
|---|---|---|---|
| Chemotherapy Treatment | |||
| Drug A | 35 (50%) | 35 (50%) | 35 (50%) |
| Drug B | 35 (50%) | 30 (42.9%) | 35 (50%) |
| Age | 47 (37, 56) | 46 (34, 56) | 45 (38, 54) |
| Unknown | <5 (<7.1%) | <5 (<7.1%) | <5 (<7.1%) |
| T Stage | |||
| T1 | 15 (21.4%) | 20 (28.6%) | 15 (21.4%) |
| T2 | 20 (28.6%) | 20 (28.6%) | 15 (21.4%) |
| T3 | 20 (28.6%) | 15 (21.4%) | 20 (28.6%) |
| T4 | 15 (21.4%) | 15 (21.4%) | 15 (21.4%) |
| 1 n (%); Median (IQR) | |||