--- title: "Tools for Education Data in R" author: "Jared E. Knowles" date: "`r Sys.Date()`" output: rmarkdown::html_vignette vignette: > %\VignetteIndexEntry{Tools for Education Data in R} %\VignetteEngine{knitr::rmarkdown} %\VignetteEncoding{UTF-8} --- # Introduction `eeptools` is an R package that seeks to make it easier for analysts at state and local education agencies to analyze and visualize their data on student, school, and district performance. By putting simple wrappers around a number of R functions, `eeptools` strives to make many common tasks simpler and less prone to error specific to analysis of education data. ```{r setup, echo = FALSE, message=FALSE, warning=FALSE, results='hide'} knitr::opts_chunk$set( cache=FALSE, comment="#>", collapse=TRUE, echo=TRUE ) library(knitr); library(eeptools) ``` # Administrative Data Functions For analysts using unit-record data of some type, there are several `calc` functions which automate common tasks including calculating ages (`age_calc`), grade retention (`retained_calc`), and student mobility (`moves_calc`). ```{r} age_calc(dob = as.Date('1995-01-15'), enddate = as.Date('2003-02-16'), units = "years") age_calc(dob = as.Date('1995-01-15'), enddate = as.Date('2003-02-16'), units = "months") age_calc(dob = as.Date('1995-01-15'), enddate = as.Date('2003-02-16'), units = "days") ``` `age_calc` also now properly accounts for leap years and leap seconds by default. `age_calc` can be passed a vector of dates of birth and a vector of end dates or a single end-date and produce a vector of ages as well -- suitable for computing student age on the fly from date-of-birth records. `retained_calc` takes a vector of student identifiers and a vector of grades and checks whether or not the student was retained in the grade level specified by the user. It returns a data.frame of all students who could have been retained and a yes or no indicator of whether they were retained. ```{r} x <- data.frame(sid = c(101, 101, 102, 103, 103, 103, 104, 105, 105, 106, 106), grade = c(9, 10, 9, 9, 9, 10, 10, 8, 9, 7, 7)) retained_calc(df = x, sid = "sid", grade = "grade", grade_val = 9) ``` `retained_calc` is intended to be used after you have processed your data as it does not take into account time or sequence other than the order in which the data is passed to it. `moves_calc` is intended to identify based on enrollment dates whether a student experienced a school move within a school year. ```{r} df <- data.frame(sid = c(rep(1,3), rep(2,4), 3, rep(4,2)), schid = c(1, 2, 2, 2, 3, 1, 1, 1, 3, 1), enroll_date = as.Date(c('2004-08-26', '2004-10-01', '2005-05-01', '2004-09-01', '2004-11-03', '2005-01-11', '2005-04-02', '2004-09-26', '2004-09-01','2005-02-02'), format='%Y-%m-%d'), exit_date = as.Date(c('2004-08-26', '2005-04-10', '2005-06-15', '2004-11-02', '2005-01-10', '2005-03-01', '2005-06-15', '2005-05-30', NA, '2005-06-15'), format='%Y-%m-%d')) moves <- moves_calc(df, sid = "sid", schid = "schid", enroll_date = "enroll_date", exit_date = "exit_date") moves ``` # Manipulate Data Another set of key functions in the package are to make basic data manipulation easier. One thing users of other statistical packaegs may miss when using R is a convenient function for determining the `mode` of a vector. The `statamode` function is designed to do just that. `statamode` works with numeric, character, and factor data types. It also includes various options for how to deal with a tie demonstrated below. ```{r statamode} vecA <- c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10) statamode(vecA, method = "stata") vecB <- c(1, 1, 1, 3:10) statamode(vecB, method = "last") vecC <- c(1, 1, 1, NA, NA, 5:10) statamode(vecC, method = "last") vecA <- c(LETTERS[1:10]); vecA <- factor(vecA) statamode(vecA, method = "last") vecB <- c("A", "A", "A", LETTERS[3:10]); vecB <- factor(vecB) statamode(vecB, method = "last") vecA <- c(LETTERS[1:10]) statamode(vecA, method = "sample") vecB <- c("A", "A", "A", LETTERS[3:10]) statamode(vecB, method = "stata") vecC <- c("A", "A", "A", NA, NA, LETTERS[5:10]) statamode(vecC, method = "stata") ``` There are a number of functions to save you keystrokes like `defac` for converting a factor to a character, `makenum` for turning a factor variable into a numeric variable, `max_mis` for taking the maximum of a vector of numerics and ignoring any NAs (useful for inclusion in `do.call` or `apply` constructions). `remove_char` allows you to quickly `gsub` out a specific character from a string vector such as an `*` or `...`. `decomma` is a somewhat specialized version of this for processing data where numerics are written with commas. `nth_max` allows you to identify the 2nd, 3rd, etc. maximum value in a vector. # Regression Models `eeptools` includes ways to simplify the use of regression analyses tools recommended by Gelman and Hill 2006 through the `gelmansim` function, which itself is a wrapper for the `arm::sim()` function. This function allows the distribution of predicted values to be generated automatically which is useful for gauging uncertainty in a statistical model and also to compare predictions from multiple models on the same case data to see if the values of those models overlap or are distinct from one another. ```{r sim} library(MASS) #Examples of "sim" set.seed (1) J <- 15 n <- J*(J+1)/2 group <- rep (1:J, 1:J) mu.a <- 5 sigma.a <- 2 a <- rnorm (J, mu.a, sigma.a) b <- -3 x <- rnorm (n, 2, 1) sigma.y <- 6 y <- rnorm (n, a[group] + b*x, sigma.y) u <- runif (J, 0, 3) dat <- cbind (y, x, group) # Linear regression dat <- as.data.frame(dat) dat$group <- factor(dat$group) M3 <- glm (y ~ x + group, data=dat) cases <- expand.grid(x = seq(-2, 2, by=0.1), group=seq(1, 14, by=2)) cases$group <- factor(cases$group) sim.results <- gelmansim(mod = M3, newdata = cases, n.sims=200, na.omit=TRUE) head(sim.results) ``` # Plotting Functions There is also a `ggplot2` version of `plot.lm` included: ```{r lmautoplot} data(mpg) mymod <- lm(cty~displ + cyl + drv, data=mpg) autoplot(mymod) ``` Finally, there is a convenient method for creating labeled mosaic plots. ```{r crossplot} sampDat <- data.frame(cbind(x=seq(1,3,by=1), y=sample(LETTERS[6:8], 60, replace=TRUE)), fac=sample(LETTERS[1:4], 60, replace=TRUE)) varnames<-c('Quality','Grade') crosstabplot(sampDat, "y", "fac", varnames = varnames, label = TRUE, title = "Crosstab Plot", shade = FALSE) ``` And without labels: ```{r} crosstabplot(sampDat, "y", "fac", varnames = varnames, label = FALSE, title = "Crosstab Plot", shade = TRUE) ``` # Datasets `eeptools` provides three new datasets of interest to education researchers. These datasets are also used in the [R Bootcamp for Education Analysts](https://www.jaredknowles.com/r-bootcamp) ```{r} library(eeptools) data("stuatt") head(stuatt) ``` The `stuatt`, student attributes, dataset is provided from the [Strategic Data Project Toolkit for Effective Data Use](https://sdp.cepr.harvard.edu/toolkit-effective-data-use). This dataset is useful for learning how to clean data in R and how to aggregate and summarize individual unit-record data into group-level data. ```{r} data(stulevel) head(stulevel) ``` The `stulevel` dataset is a simulated student-level longitudinal record. It contains student and school level attributes and is useful for practicing evaluating longitudinal analyses of student unit-record data. ```{r} data("midsch") head(midsch) ``` The `midsch` dataset contains an analysis on abnormality in school average assessment scores. It contains observed and predicted values of aggregated test scores at the school level for a large midwestern state.