My presentation at RStudio::conf(2019), Austin, Tx - "The lazy and easily distracted report writer" on 18th January 2018.
The example provided shows how you can use parameters with rmarkdown and knitr to produce reports for multiple outcomes.
We're using a parameter passed into the
report via the YAML header. This parameter can be set at time of rendering the
report (using the knit with parameters option in the RStudio IDE,
via "Inputs" specification in RStudio Connect or by specifying the parameters
and values within a list passed to the params argument of rmarkdown:render.
If the user sets the quantAudience parameter to FALSE then the code is hidden
in the rendered HTML, and certain text sections are hidden. To hide text we
keep the relevant text in child documents and only bring these into the report
if the quantAudience parameter is set to TRUE. The code is entirely contained
in the parent document, but is executed depending on the setting of the parameter.
The trick is to rename the outcome variables (and potentially associated
variables with the same name e.g. HAMDTL17_BASE, HAMDTL17_CHG) to something
generic i.e. outcome that will allow us to write code once and use it
for each of the endpoint variables. We can then use the endpoint parameter
params$endpoint to specify labels and names wherever these are needed. We
could also compute based on the value of params$endpoint if we need to do
specific things for each endpoint e.g. if one outcome is binary, rather than
continuous.