Tools for Education Data in R

Jared E. Knowles

2020-05-02

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.

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).

age_calc(dob = as.Date('1995-01-15'), enddate = as.Date('2003-02-16'), 
         units = "years")
#> [1] 8.087671
age_calc(dob = as.Date('1995-01-15'), enddate = as.Date('2003-02-16'), 
         units = "months")
#> [1] 97.03571
age_calc(dob = as.Date('1995-01-15'), enddate = as.Date('2003-02-16'), 
         units = "days")
#> Time difference of 2954 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.

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)
#>   sid retained
#> 1 101        N
#> 2 102        N
#> 3 103        Y
#> 4 105        N

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.

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
#>   sid moves
#> 1   1     4
#> 2   2     4
#> 3   3     2
#> 4   4    NA

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.

vecA <- c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10)
statamode(vecA, method = "stata")
#> [1] "."
vecB <- c(1, 1, 1, 3:10)
statamode(vecB, method = "last")
#> [1] 1
vecC <- c(1, 1, 1, NA, NA, 5:10)
statamode(vecC, method = "last")
#> [1] 1
vecA <- c(LETTERS[1:10]); vecA <- factor(vecA)
statamode(vecA, method = "last")
#> [1] J
#> Levels: J
vecB <- c("A", "A", "A", LETTERS[3:10]); vecB <- factor(vecB)
statamode(vecB, method = "last")
#> [1] A
#> Levels: A
vecA <- c(LETTERS[1:10])
statamode(vecA, method = "sample")
#> [1] "I"
vecB <- c("A", "A", "A", LETTERS[3:10])
statamode(vecB, method = "stata")
#> [1] "A"
vecC <- c("A", "A", "A", NA, NA, LETTERS[5:10])
statamode(vecC, method = "stata")
#> [1] "A"

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.

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)
#>      x group      yhats   yhatMin  yhatMax
#> 1 -2.0     1  1.1736195 -6.264184 8.243267
#> 2 -1.9     1  0.7390300 -7.376271 8.587548
#> 3 -1.8     1  1.1866869 -6.256829 7.846494
#> 4 -1.7     1 -0.3616534 -8.355161 7.494966
#> 5 -1.6     1  0.1931550 -7.104866 8.648023
#> 6 -1.5     1 -0.6293359 -7.803899 6.654324

Plotting Functions

There is also a ggplot2 version of plot.lm included:

data(mpg)
mymod <- lm(cty~displ + cyl + drv, data=mpg)
autoplot(mymod)
#> `geom_smooth()` using formula 'y ~ x'
#> `geom_smooth()` using formula 'y ~ x'
#> `geom_smooth()` using formula 'y ~ x'
#> `geom_smooth()` using formula 'y ~ x'

Finally, there is a convenient method for creating labeled mosaic plots.

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:

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

library(eeptools)
data("stuatt")
head(stuatt)
#>   sid school_year male race_ethnicity birth_date first_9th_school_year_reported
#> 1   1        2004    1              B      10869                           2004
#> 2   1        2005    1              H      10869                           2004
#> 3   1        2006    1              H      10869                           2004
#> 4   1        2007    1              H      10869                           2004
#> 5   2        2006    0              W      11948                             NA
#> 6   2        2007    0              B      11948                             NA
#>   hs_diploma      hs_diploma_type hs_diploma_date
#> 1          0                                     
#> 2          0                                     
#> 3          0                                     
#> 4          0                                     
#> 5          1     Standard Diploma        6/5/2008
#> 6          1 College Prep Diploma       5/24/2009

The stuatt, student attributes, dataset is provided from the Strategic Data Project Toolkit for 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.

data(stulevel)
head(stulevel)
#>     X school  stuid grade schid dist white black hisp indian asian econ female
#> 1  44      1 149995     3   495  105     0     1    0      0     0    0      0
#> 2  53      1  13495     3   495   45     0     1    0      0     0    1      0
#> 3 116      1 106495     3   495   45     0     1    0      0     0    1      0
#> 4 244      1  45205     3   205   15     0     1    0      0     0    1      0
#> 5 274      1 142705     3   205   75     0     1    0      0     0    1      0
#> 6 276      1  14995     3   495  105     0     1    0      0     0    1      0
#>   ell disab sch_fay dist_fay luck   ability    measerr      teachq year attday
#> 1   0     0       0        0    0  87.85405  11.133264 39.09024712 2000    180
#> 2   0     0       0        0    1  97.78756   6.822394  0.09848192 2000    180
#> 3   0     0       0        0    0 104.49303  -7.856159 39.53885270 2000    160
#> 4   0     0       0        0    1 111.67151 -17.574152 24.11612277 2000    168
#> 5   0     0       0        0    0  81.92539  52.983338 56.68061304 2000    156
#> 6   0     0       0        0    0 101.92904  22.604145 71.62196655 2000    157
#>   schoolscore district schoolhigh schoolavg schoollow   readSS   mathSS
#> 1    29.22427        3          0         1         0 357.2865 387.2803
#> 2    55.96326        3          0         1         0 263.9046 302.5724
#> 3    55.96326        3          0         1         0 369.6722 365.4614
#> 4    55.96326        3          0         1         0 346.5957 344.4964
#> 5    55.96326        3          0         1         0 373.1254 441.1581
#> 6    55.96326        3          0         1         0 436.7607 463.4033
#>       proflvl race
#> 1       basic    B
#> 2 below basic    B
#> 3       basic    B
#> 4       basic    B
#> 5       basic    B
#> 6  proficient    B

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.

data("midsch")
head(midsch)
#>   district_id school_id subject grade n1   ss1 n2   ss2 predicted  residuals
#> 1          14       130    math     4 44 433.1 40 463.0  468.7446 -5.7445937
#> 2          70        20    math     4 18 443.0 20 477.2  476.4765  0.7235053
#> 3         112        80    math     4 86 445.4 94 472.6  478.3509 -5.7508949
#> 4         119        50    math     4 95 427.1 94 460.7  464.0586 -3.3585931
#> 5         147        60    math     4 27 424.2 27 458.7  461.7937 -3.0936928
#> 6         147       125    math     4 17 423.5 26 463.1  461.2470  1.8530072
#>       resid_z     resid_t       cooks test_year     tprob flagged_t95
#> 1 -0.59189645 -0.59170988 0.000171271      2007 0.2787298           0
#> 2  0.07455731  0.07452135 0.000003510      2007 0.4706873           0
#> 3 -0.59266905 -0.59248250 0.000244921      2007 0.2774827           0
#> 4 -0.34605798 -0.34591020 0.000059900      2007 0.3650957           0
#> 5 -0.31877383 -0.31863490 0.000054100      2007 0.3762745           0
#> 6  0.19093568  0.19084643 0.000019800      2007 0.4250936           0

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.