In softloud/sperm: Analysis

knitr::opts_chunk$set(echo = FALSE)

# Learn more about creating websites with Distill at:
# https://rstudio.github.io/distill/website.html

# Learn more about publishing to GitHub Pages at:
# https://rstudio.github.io/distill/publish_website.html#github-pages

library(tidyverse)
library(sperm)
library(targets)
library(gt)

# packages used in this readme
library(tidyverse)
library(sperm)
library(targets)
library(gt)

Installation

The sperm:: analysis is available at GitHub with:

Either download the codebase (green Code button) as a .zip file or use GitHub command-line interface (gh cli)

gh repo clone softloud/sperm

The raw data are preloaded in the packaged analysis which can be accessed by installing the sperm:: package:

# install.packages("devtools")
devtools::install_github("softloud/sperm")

sperm:: is created using targets:: so that each step in the analysis can be accessed.

install.packages("targets")

sperm::'s analysis pipeline can be viewed:

tar_glimpse()

sperm:: uses the renv:: package to snapshot the versions of packages used to ensure reproducibility, run this command to load the required packages after opening the R project.

renv::restore()

Raw data

The raw data was last downloaded:

file.info("data/count_obs.rda") %>% pull(ctime)

There are four raw datasets preloaded in sperm:: package: count_obs, morphology_obs, motility_obs, and volumne_obs scraped from the googlesheet.

count_obs

Cleaned data

The cleaned data can be accessed using targets:

tar_read(model_dat)

Modelling variables

We're interested in assessing

outcome = intervention_class + type of infertility

Currently class is set to major_intervention_grouping and intervention is set to grouped_intervention. The moderator is type_of_infertility.

Kerry, is this correct? Very easy to update if not :)

Cleaning class labels

There are 10 interventions with more than one class label.

tar_read(qa_class) %>% 
  gt()

There needs to be at most one class label per intervention or the model will complain. Charles has set the class to be the most-used class label for each intervention.

Kerry please check you're happy with the class labels :)

tar_read(model_dat) %>% 
  count(intervention, class) %>% 
  select(-n) %>% 
  gt()

Modelling input

Here's a random sample of ten observations:

tar_read(wide_obs) %>% 
  sample_n(10) %>% 
  gt() %>% 
  tab_header("Modelling data",
             subtitle = "10 randomly-selected observations") %>% 
  tab_options(
  )