In tntp/tntpr: Data Analysis Tools Customized for TNTP

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(tntpr)
library(tidyverse)

About

The tntpr package makes data science at TNTP easier and more accurate by supplying tools that are needed for common TNTP analyses.

Package summary

Some of the highlights of the package include:

A TNTP-themed RMarkdown template, for starting a new analysis with a shell that can already generate a TNTP-themed .docx report
Functions for initializing a new repository or project folder with TNTP-standard directories and documentation
TNTP-specific ggplot2 themes and color palettes
Survey analysis tools
Wrappers for quickly making typical TNTP-style charts (e.g., bar chart of the distribution of one variable, grouped by a second)
Education-specific data management functions (e.g., date_to_SY() to convert continuous hire dates into school years using a specified cutoff date), and a built-in fake student achievement dataset to play with called wisc.

Installing the package

This package is not on CRAN, and probably will not ever be. You'll need to install this package from its GitHub repository. You can add this to the top of your analysis script:

library(tntpr)
library(magrittr)
library(ggplot2)
library(tidyr)
library(dplyr)

Once installed, you can update the package with update_tntpr().

Usage

Reporting templates

Start your analysis with a good-looking .docx file as output, right off the bat. Right now we have just a single TNTP template, "Data Memo", but it can be adapted and improved and if we have other common needs (a different set of headings?) those can easily be separate templates.

To access templates once you've installed the tntpr package: go to File -> New File -> R Markdown -> From Template. You'll see a choice "Data Memo" from the tntpr package. Just specify the document's file name and the directory you want it in (probably a subfolder of a Bitbucket repository) and you're off!

A file tntp-style-file.docx will be copied into that directory; leave it there. That provides the TNTP .docx stylings when you re-knit your R Markdown document.

Setting up repositories and subfolders

This saves time getting started with a new analysis and encourages use of common file storage conventions and documentation that make work more browsable and transparent.

Directory conventions

TNTP's Bitbucket directory structure is a single repository per client or region, with subdirectories in the repository corresponding to specific analysis projects. For example:

...

Usage

After creating a new, empty repo in Bitbucket and cloning it, run setup_repo to initialize it. This will create a subfolder as well, in which you'll conduct an analysis project.

If the repository already exists, and you just want to begin a new analysis project, create the new subdirectory with setup_subdirectory.

Both functions take the same arguments, used to setup the project subfolder and its README:
- subfolder: what the name of the subdirectory should be, e.g., "xyz_instructional_audit"
- proj_name: the full name of the analysis project, e.g., "XYZ Public Schools Equity Study". Appears in the README.
- analyst_name: the analyst(s) working on this project. Appears in the README.

setup_repo will also add a .Rproj RProject file, .gitignore file, and create a README.Md file for the main repository.

TNTP colors

You can access the official TNTP-branded colors using palette_tntp(). This will return a vector with hex code for our colors:

palette_tntp("dark_blue", "orange", "light_gray")

Or you can select a specific TNTP palette ("default", "colors_tntp_classic", "likert_4pt", "likert_5pt", or "likert_6pt") with palette_tntp_scales and return a vector with hex codes for that TNTP palette.

palette_tntp_scales(palette = "likert_5pt")

You can use these scale palettes as fill or color aesthetics in ggplot with scale_fill_tntp and scale_color_tntp.

data.frame(
  question = "To what extent do you agree...",
  response = c(
    rep("Strongly disagree", 3),
    rep("Disagree", 4),
    rep("Somewhat disagree", 3),
    rep("Somewhat agree", 4),
    rep("Agree", 10),
    rep("Strongly agree", 2)
  )
) %>%
  mutate(response = response %>% factor(levels = rev(c(
    "Strongly disagree",
    "Disagree",
    "Somewhat disagree",
    "Somewhat agree",
    "Agree",
    "Strongly agree"
  )))) %>%
  ggplot(aes(question, fill = response)) +
  geom_bar(position = position_fill()) +
  theme_tntp_2018(axis_text = "Y", grid = FALSE) +
  labs(
    x = NULL, y = NULL,
    fill = "Response"
  ) +
  coord_flip() +
  scale_fill_tntp(palette = "likert_6pt")