Skip to content
reportifyr documentation

Performing Analyses

Examining a Simple Workflow

Let’s first create a concentration-time plot of the Theophylline data set:

library(reportifyr)
library(ggplot2)
library(dplyr)


data <- Theoph


p <- ggplot(data, aes(x = Time, y = conc, group = Subject)) +
  geom_point() +
  geom_line() +
  theme_bw() +
  labs(x = "Time (hr)", y = "Theophylline concentration (mg/L)")


p

Saving the Figure and Creating Metadata

If we want to include this plot into the report we’ll need to save this figure to the OUTPUTS/figures subdirectory. There are two separate processes to accomplish this - the easiest way is to call the wrapper function ggsave_with_metadata():

figures_path <- here::here("OUTPUTS", "figures")
plot_file_name <- "theoph-pk-plot.png"


ggsave_with_metadata(
  filename = file.path(figures_path, plot_file_name),
  plot = p,
  width = 6,
  height = 4
)

Alternatively you could call ggplot2::ggsave() and then call write_object_metadata():

ggplot2::ggsave(
  filename = file.path(figures_path, plot_file_name),
  plot = p,
  width = 6,
  height = 4
)


write_object_metadata(object_file = file.path(figures_path, plot_file_name))

Both processes give the same final result, however, severing the tie between saving a figure/table (artifact) and writing its metadata may lead to the these artifacts becoming out of sync.

Using Metadata to Standardize Footnotes

This brings us to metadata. reportifyr uses metadata to create a record of the artifact being inserted into the Microsoft Word document (report) to aid in reproducibility and traceability.

Specifically, reportifyr uses a meta_type parameter to inject a portion of standardized footnotes into the report. We can use the function get_meta_type() to see what artifact meta_type values are currently saved in the standard_footnotes.yaml within the ‘report’ subdirectory upon initialization. We just need to provide the path to the standard_footnotes.yaml that we want to pull meta_type from.

Below are the basic meta_type included automatically within reportifyr. If a new meta_type is required for your report, simply follow the formatting style in the standard_footnotes.yaml file to expand the list.

meta_type <- get_meta_type(here::here("report", "standard_footnotes.yaml"))


names(meta_type)
[1] "logistic_plot"          "boxplot"                "TTE_plot"               "correlation_plot"       "GOF_plot"               "ETA_histogram"          "VPC"                    "kaplan-meier-plot"      "conc-time-trajectories" "linear-regression-plot" "probit-regression-plot"
[12] "univariate"             "final"                  "full"                   "alternative"            "univariate_TTE"         "final_TTE"              "full_TTE"               "alternative_TTE"        "ER_summary"             "covariate_descriptive"  "frequency"
[23] "probit-regression-fit"  "linear-regression-fit"  "cox-regression-fit"

Now, let’s re-save the plot object with the conc-time-trajectories meta_type provided:

plot_file_name_meta <- "theoph-pk-plot-meta.png"


ggsave_with_metadata(
  filename = file.path(figures_path, plot_file_name_meta),
  plot = p,
  meta_type = meta_type$`conc-time-trajectories`,
  width = 6,
  height = 4
)

Using meta_type$ also allows for tab-completion to help you select the correct meta_type! Let’s use the preview_metadata_files() to see the difference when using a meta_type:

metadata <- preview_metadata_files(file_dir = figures_path)


knitr::kable(metadata)

From the data frame, we can see that there are additional parameters that can be saved as part of an artifact’s metadata using reportifyr’s wrapper functions (i.e., meta_abbreviations, meta_equations, and meta_notes).

Similar to meta_type, we can retrieve a list of standardized meta_abbrevs stored in the standard_footnotes.yaml by using get_meta_abbrevs():

meta_abbrevs <- get_meta_abbrevs(here::here("report", "standard_footnotes.yaml"))


names(meta_abbrevs)
[1] "AUC"       "AGEBL"     "ALBBL"     "ALTBL"     "ASTBL"     "B2MBL"     "BSABL"     "BSTGBGE2"  "BSTGEQ3"   "CADC_21D"  "CADC_28D"  "CADC_42D"  "CADC_56D"  "CATABL"    "CAVGA_21D" "CAVGA_28D" "CAVGA_42D" "CAVGA_56D" "CAVGM_21D" "CAVGM_28D" "CMAXA"     "CMAXM"     "CNCEURO"   "CRPBL"
[25] "CYTORSK"   "ECOGBEQ2"  "ECOGBGE1"  "GLAUCBL"   "HEPFCH"    "HISTDIAB"  "HSDRYEYE"  "HSIOSURG"  "IBMIBL"    "IGGBL"     "IGGFLG"    "KERABL"    "LDHBL"     "LENSTABL"  "MEDFL"     "MMIGTY"    "MMLCTY"    "MMTY"      "MPROTBL"   "PCD38"     "PRXGT3"    "PRXGT4"    "RACEBFLG"  "RACEWFLG"
[49] "RENALFCH"  "SBCMABL"   "SCH"       "SCHCH"     "SEXN"      "SPLITB"    "STAGEBL"   "STRETCHB"  "TBILBL"    "WTBL"

Now, instead of saving the second figure (plot_file_name_meta) out again using gg_save_with_metadata() and providing the additional arguments, let’s instead use update_object_footnotes() to add a mock meta_abbrevs:

update_object_footnotes(
  file_path = file.path(figures_path, plot_file_name_meta),
  overwrite = TRUE,
  meta_abbrevs = c(meta_abbrevs$CMAXA)
)

Once we have updated the footnotes for the second figure, let’s take one last preview of its metadata to confirm our changes.

metadata_file <- preview_metadata(file_name = file.path(figures_path, plot_file_name_meta))


knitr::kable(metadata_file)

Performing an Analyses

With a simple workflow examined, let’s perform an analysis computing subject level pharmacokinetic parameters and then study level statistics of the data set to generate an additional plot. We’ll compute maximum drug concentration (CMAX), time to peak drug concentration (TMAX), and area under the concentration-time curve (AUC) for the Theophylline data set before performing a linear regression between baseline weight (WTBL) and AUC.

Generating a Figure for Use with reportifyr

calc_auc_linear_log <- function(time, conc) {
  auc <- 0


  cmax_index <- which.max(conc)


  for (i in 1:(length(time) - 1)) {
    delta_t <- time[i + 1] - time[i]


    if (i < cmax_index) {


      auc <- auc + delta_t * (conc[i + 1] + conc[i]) / 2
    } else if (i >= cmax_index && conc[i + 1] > 0 && conc[i] > 0) {


      auc <- auc + delta_t * (conc[i] - conc[i + 1]) / log(conc[i] / conc[i + 1])
    } else {


      auc <- auc + delta_t * (conc[i + 1] + conc[i]) / 2
    }
  }


  return(auc)
}


pk_params <- data %>%
  mutate(Subject = as.numeric(Subject)) %>%
  group_by(Subject) %>%
  summarise(
    cmax = max(conc, na.rm = TRUE),
    tmax = Time[which.max(conc)],
    auc = calc_auc_linear_log(Time, conc),
    wt = Wt %>% unique()
  )


knitr::kable(pk_params)

lr <- pk_params %>%
  ggplot(aes(x = wt, y = auc)) +
  geom_point() +
  geom_smooth(method = "lm", formula = y ~ x, se = TRUE, color = "blue") +
  theme_bw() +
  labs(x = "Subject weight (kg)", y = "AUC (hr mg/L)")


lr

Let’s save this plot with the linear-regression-plot meta_type and the relevant AUC meta_abbrevs.

plot_file_name <- "theoph-pk-exposure.png"


ggsave_with_metadata(
  filename = file.path(figures_path, plot_file_name),
  plot = lr,
  meta_type = meta_type$`linear-regression-plot`,
  meta_abbrevs = c(meta_abbrevs$AUC),
  width = 6,
  height = 4
)

Generating a Table for Use with reportifyr

With our linear regression plot generated and saved, let’s now save out the pk_params data frame to .csv using write_csv_with_metadata() so we can include it in the report. We can pass in similar meta_type, meta_notes, and meta_abbrevs arguments, if applicable, as well as any arguments that would be used in write.csv().

For this table, let’s use row.names = FALSE:

tables_path <- here::here("OUTPUTS", "tables")
outfile_name <- "theoph-pk-parameters.csv"


write_csv_with_metadata(
  object = pk_params,
  file = file.path(tables_path, outfile_name),
  meta_abbrevs = c(meta_abbrevs$AUC),
  row.names = FALSE
)

Similarly, let’s also save out the Theophylline data set to include in the report. This time we will format the table using format_flextable() before saving the table out as an .RDS file:

library(flextable)


data_outfile_name <- "theoph-pk-data.RDS"


table <- format_flextable(
  data_in = data,
  table1format = FALSE
)


save_rds_with_metadata(
  object = table,
  file = file.path(tables_path, data_outfile_name)
)

At the end, we’ve generated two figures and two tables and saved them alongside their metadata using reportifyr, allowing us to easily incorporate them into our upcoming Drafting Reports!