philsf / philsfmisc Goto Github PK

Um template de relatório para documentar todas as escolhas metodológicas, resultados, figuras e conclusões da análise de dados, conforme as lições aprendidas com documentações na indústria

fix a warning related to a ggplot2 function that was not properly imported.

Some styles were not configured in the 0.6.1 release:

• Title (Arial 18, bold)
• Subtitle (Arial 12, bold)
• Author (Arial 11)
• Date (Arial 10)

Automatic paragraphs before/after each of these styles? Current settings:

• Title 0.85cm / 0.42 cm
• Subtitle 0.42 cm / 0.42 cm
• Author/Date 0cm / 0cm

Create a gtsummary theme that

• prints % symbol in tables
• uses mean/SD + t.test by default
• is consistent with survey objects (mean/SD + t.test)
• uses survey adjusted chi square by default
• uses "verbose" CIs (ll to ul)
• uses 3 digits p-values

Revert the default of pct.symbol to TRUE.

Create a new Rmarkdown template to create Statistical Analysis Plans (SAP) for future analyses.

Use the same formatting choices as the report template.

All geo* functions assume base 10. It would be nice if there was an argument to use more general bases

All format.* functions must be changed into format_* so the build check passes.

This, in turn, will require that all packages that depend on philsfmisc be updated accordingly.

Perhaps we can provide aliases to ease this upgrading process.

• rename functions
• update README.md

We are beginning to replicate the formatting of OR results so it would be convenient to have a handy function for that.

Include gg theme in a function. How about theme_ff()?

Relates to philsf/philsf.workflow#106

Include the function used to compute the geometric mean, in base 10.

The styles doc is undergoing a major overhaul, that will enhance

• Standard font style
• Section styles
• Table display as centered
• Table format style

As soon as this styles.docx is finished, incorporate this enhancement in the report template used for consulting reports.

• Remove templates
• Remove README template section

philsf/philsf.workflow#2

Convert manually created docs to roxygen comments inside each source file.

This will help future iterations of development with the automation tools available in devtools and RStudio.

Not every situation we are interested in printing perecentages with the percent sign.

This should be an optional (logical) argument. In fact, the default should be FALSE.

The final template for the reports has too narrow margins.

Implement a function to compute the geometric SD

10^sd(log10(data))

The current state of affairs prevents the use of format.pct in format.interval, because format.float returns "NA" for both ends of the interval.

Possible solutions

• include an argument in format.interval to consider the interval as proportions
• include a check in format.float to see if the last char is "%"

Provide a new function (format.pct?) to format proportion 0 <= p <= 1 into a percentage.

Consider defaulting to 0 digits.

Several documentation warnings are appearing in new builds:

• argument name mismatches between code and doc
  • format.float
  • cv2logsd
  • cv2logvar
  • geocv
  • geomean
  • geosd
  • logsd2cv
  • logvar2cv
  • predint
• error in format.float example (F instead of FALSE)
• error in format.interval example (intentional, but must not run)

Several new functions were created, they must be, at minimum, be displayed in the README

format.float needs at least a minimum documentation page for the package.

Similar to logsd2cv, there could be a function to convert var -> cv

Need to re-structure the styles.docx, for use in RStudio.

Create a function is.interval(x) to check if

• x is a vector of 2 values
• each of the values is.numeric()
• the second value is greater or equal than the first value

Expected return value is.logical().

Currently the analytical dataset mockup is being created in a "manual" fashion.
Move this code to a new function

• output is a dataframe/tibble with the same shape, but id goes from 1:N
• optional named id col (with a reasonable default "id")
• "N" argument: boolean for whether or not to expose the actual N or print "N"

The report template does not have a placeholder signagture field.

We should include one by default. In case it will not be used, it can be removed afterwards.

Create a function to compute the CV of log-normal data, given the SD of log-transformed data.

sqrt(exp(lSD ^ 2) - 1)*100

where lSD is sd(log(data))

Note: come up with a descriptive function name!

At least geomean has a poorly written examples section. Fix those for release.

This repo needs a bare minimum README with

• package description
• installation instructions

The solution in #49 should not have been implemented in format.interval, but in format.float for greater impact.

We are beginning to replicate the formatting of OR results so it would be convenient to have a handy function for that.

• There are empty description in the following docs
  • cv2logsd
  • cv2logvar
  • geocv
  • geomean
  • is.within
  • logsd2cv
  • logvar2cv

format.interval needs at least a minimum documentation page for the package.

Include the function to compute the geometric CV of a log10-distributed vector of values.

Several applications require the computation of a PI of data, with a given significance value.

We should provide a function that implements the most common formula.

The standard in R is to provide proportions, and not percentages. Since we can now pretty-format percentages ( #30 ), we can change the API to consistently produce/deal with proportions in all CV functions.

There are currently no unit tests for this package. Use testthat to test each function with some predictable results.

is.within needs at least a minimum documentation page for the package.