dth2 / undiagnosed Goto Github PK

Updated 1/19/15

####More urgent

• Split files in HIVBackCalc/R by function, for helping developers. Use nice section headings within those .R files
• Embed non-proprietary datasets to use with tutorials, e.g. simulated KC data
• Create vignette
• Release minimal package to accompany PLoS One paper
• Update Shiny to use this package

####Less urgent

• Map out code terminology - in the "Index to functions" below
• Improve terminology
  • "infPeriod" variable -> "tid"
  • "fig1" function -> "tidSurvivorFxn" ?
  • "impute" -> includeMiss vs excludeMiss ?
  • Learn about search and replace across multiple files
  • Test code and make sure it works with updated terminology
• Turn scripts (e.g., format_data.R) into functions that can be embedded in the package and essentially serve as a wrapper for running a sequence of functions?
• Get Sam's input on
  • Function-ifying scripts
  • When to use Markdown and HTMl vs Sweave and PDF formats
  • Pros/cons of eventually migrating to using S2, S3 or S4 classes
• Add two modes to setup_hivbackcalc: package updated = TRUE (just load HIVBackCalc package) vs FALSE (need to source the package files and source the sim'd data) ?

Structure is fairly obvious: the data formatting file is sourced by the run file, so the run file is self-contained.

Source "run_JKB.R" or knit "replication_JKB.Rmd". You should get the results reported in the submitted paper. Both these files are basically Ian's code from run.R, but with section headers and comments to clarify the code.

####Guide to .R code

Note on knitr hooks and code chunks

The .R files in this folder contain knitr "hooks" that define code "chunks". These "chunks" are read in at the beginning of reports and then executed at subsequent points in the report.

The advantage of this approach is that it makes it easy to to have R code that you can execute by section throughout a report, as opposed to executing it all at once by sourcing the whole .R file. But it's really convenient that the code is also fully contained and maintainable through a standalone .R file that can be sourced all together, when desired.

The process is:

1. Insert "hooks" into R code to define "chunks" of code. Hook syntax is

## ---- hookname ----

2. At the beginning of your .Rnw or .Rmd report file, after you load the knitr library, use the read_chunk() function to read in the R code, e.g.

read_chunk('path_to_file.R')

3. Throughout the report, execute code chunks using the hook name, e.g.

<<hookname, echo=FALSE>>=
@

####Guide to analyses stored in this folder

####Guide to file types

1. Download HIVBackCalc_1.01.tar.gz (Unix) or HIVBackCalc_1.01.zip (Windows) from this repository, and install the package using the following instructions:

# Open R in/set the R workding directory to the folder containing the downloaded zip file and type:
# (switch in .tar.gz for .zip if on Unix)
install.packages('HIVBackCalc_1.01.zip', repos=NULL)
library(HIVBackCalc)

2. Identify the .Rnw file for the appropriate analysis
3. In the setup_hivbackcalc() function call(s) near the top of the report, change the working directory to reflect yours.
4. Load the knitr library in R
5. Use the purl() function from the knitr library to run only the R code without generating the report. I think that if you use the tangle=TRUE option, you will get an .R file of all the code in the report. You could then execute code from that .R file.

####Estimation Functions

####Formatting Functions

####Plotting Functions