Plotting with a <summarised_result>
For the purpose of this vignette we will turn summarise the dataset
penguins from the palmerpenguins package using
PatientProfiles::summariseResult()
library(PatientProfiles)
library(palmerpenguins)
library(dplyr)
#>
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#>
#> filter, lag
#> The following objects are masked from 'package:base':
#>
#> intersect, setdiff, setequal, union
summariseIsland <- function(island) {
penguins |>
filter(.data$island == .env$island) |>
summariseResult(
group = "species",
includeOverallGroup = TRUE,
strata = list("year", "sex", c("year", "sex")),
variables = c(
"bill_length_mm", "bill_depth_mm", "flipper_length_mm", "body_mass_g",
"sex"),
estimates = c(
"median", "q25", "q75", "min", "max", "count_missing", "count",
"percentage", "density")
) |>
suppressMessages() |>
mutate(cdm_name = island)
}
penguinsSummary <- bind(
summariseIsland("Torgersen"),
summariseIsland("Biscoe"),
summariseIsland("Dream")
)Plotting principles
Although the input of a visOmopResult plot function can be a
<summarised_result> when referring to inputs, we will
use the columns shown in its tidy format:
tidyColumns(penguinsSummary)
#> [1] "cdm_name" "species" "year" "sex"
#> [5] "variable_name" "variable_level" "count" "median"
#> [9] "q25" "q75" "min" "max"
#> [13] "count_missing" "density_x" "density_y" "percentage"And not the column names seen in the
<summarised_result> format:
colnames(penguinsSummary)
#> [1] "result_id" "cdm_name" "group_name" "group_level"
#> [5] "strata_name" "strata_level" "variable_name" "variable_level"
#> [9] "estimate_name" "estimate_type" "estimate_value" "additional_name"
#> [13] "additional_level"That’s the most important thing to be aware of, otherwise populating arguments becomes a bit tricky.
We should always subset the <summarised_result> to
the variable_name of interest before calling the plotting functions. It
is not advised to combine results of different variables if they don’t
share common estimates that we want to plot as in the tidy form NAs will
be created.
Scatter plot
We can create simple scatter plots using the
plotScatter() let’s see some examples:
penguinsSummary |>
filter(variable_name == "bill_depth_mm") |>
filterStrata(year != "overall", sex == "overall") |>
scatterPlot(
x = "year",
y = "median",
line = TRUE,
point = TRUE,
ribbon = FALSE,
facet = "cdm_name",
colour = "species"
)
Additionally, we can use the function themeVisOmop() to
change the default ggplot2 style to our default style. Not
only that, but we can use standard ggplot2 functionalities to the
returned plot:
penguinsSummary |>
filter(variable_name %in% c("bill_length_mm", "bill_depth_mm"))|>
filterStrata(year == "overall", sex == "overall") |>
filterGroup(species != "overall") |>
scatterPlot(
x = "density_x",
y = "density_y",
line = TRUE,
point = FALSE,
ribbon = FALSE,
facet = cdm_name ~ variable_name,
colour = "species"
) +
themeVisOmop() +
ggplot2::facet_grid(cdm_name ~ variable_name, scales = "free_x") 
penguinsSummary |>
filter(variable_name == "flipper_length_mm") |>
filterStrata(year != "overall", sex %in% c("female", "male")) |>
scatterPlot(
x = c("year", "sex"),
y = "median",
ymin = "q25",
ymax = "q75",
line = FALSE,
point = TRUE,
ribbon = FALSE,
facet = cdm_name ~ species,
colour = "sex",
group = c("year", "sex")
) +
themeVisOmop(fontsizeRef = 10, legendPosition = "top") +
ggplot2::coord_flip() +
ggplot2::labs(y = "Flipper length (mm)")
penguinsSummary |>
filter(variable_name %in% c(
"flipper_length_mm", "bill_length_mm", "bill_depth_mm")) |>
filterStrata(sex == "overall") |>
scatterPlot(
x = "year",
y = "median",
ymin = "min",
ymax = "max",
line = FALSE,
point = TRUE,
ribbon = TRUE,
facet = cdm_name ~ species,
colour = "variable_name",
group = c("variable_name")
) +
themeVisOmop(
fontsizeRef = 10,
legendPosition = "top"
)
Bar plot
Let’s create some simple bar plots:
penguinsSummary |>
filter(variable_name == "number records") |>
filterGroup(species != "overall") |>
filterStrata(sex != "overall", year != "overall") |>
barPlot(
x = "year",
y = "count",
colour = "sex",
facet = cdm_name ~ species
) +
themeVisOmop()
Box plot
Let’s create some box plots of their body mass:
penguinsSummary |>
filter(variable_name == "body_mass_g") |>
boxPlot(x = "year", facet = c("cdm_name", "species"), colour = "sex") +
themeVisOmop(fontsizeRef = 10)
We can specify how we want to facet using a formula:
penguinsSummary |>
filter(variable_name == "body_mass_g") |>
boxPlot(x = "year", facet = cdm_name ~ species, colour = "sex") +
themeVisOmop(fontsizeRef = 10, legendPosition = "bottom")
Arguments before ~ specify the rows, and arguments after
the columns, see another example faceting also by sex in the columns.
And in this case colouring by year:
penguinsSummary |>
filter(variable_name == "body_mass_g") |>
filterGroup(species != "overall") |>
filterStrata(sex %in% c("female", "male"), year != "overall") |>
boxPlot(x = "cdm_name", facet = sex ~ species, colour = "year") +
themeVisOmop()
Note that as we didnt specify x there is no levels in the x axis, but box plots are produced anyway.
