Project 1: Exploring 2021-2022 PARCC Assessment Data
Nandini Srinivasan
2023-10-14
Project 1: Exploring 2021-2022 PARCC Data for Students with Disabilities
Introduction
The PARCC (Partnership for Assessment of Readiness for College and Careers) is an annual assessment of mathematics and ELA (English language arts/literacy) administered by the District of Columbia. Students in grades 3-12 who attend public schools and public charter schools take state assessments in these subject areas each spring. These assessments are built to assess Common Core State Standards. Students who attend private schools are not required to take these assessments.
PARCC scores range from Levels 1-5. Levels 4 and 5 indicate that the student is on track to enter the next grade level, and to graduate from high school ready for a career or college.
Level 1: Did Not Yet Meet Expectations
Level 2: Partially Met Expectations
Level 3: Approached Expectations
Level 4: Met Expectations Level
5: Exceeded Expectations
Students with disabilities may utilize testing accommodations and accessibility features as needed for PARCC. Students in grades 3-11 with significant cognitive disabilities may qualify to take an alternate assessment, MSAA (the Multi-State Alternate Assessment).
The dataset I will explore contains PARCC and MSAA participation and performance data for all students with disabilities who took these assessments during the 2021-2022 school year, as well as students without disabilities. I plan to explore whether test results vary by school/LEA (Local Education Agency), grade level, and/or subject tested. I will also explore whether the proportion of students with disabilities taking PARCC and MSAA varies by school.
Context of this dataset: 2021-2022 was the first full year of in-person learning for most K-12 students in DC. According to OSSE (Office of the State Superintendent of Education), approximately half of the students who completed a statewide assessment this year did so for the first time.
Dataset Information
Scores are reported for students who were enrolled for the full academic year and attended at least 85% of the days between Enrollment Audit Count Day (October 5, 2021) and the first day of the state assessment window (March 14, 2022 for MSAA and April 4, 2022 for PARCC).
The required assessments for high school are Algebra 1, Geometry, English I, English II, and Algebra II (for students who take Geometry prior to high school). Middle school students who take high school mathematics assessments are counted for performance reporting.
At least 10 students in a school and/or a specific group of students must have taken the test in order for performance to be aggregated. Schools/groups that did not meet this threshold received a value of “N<10”.
Recently arrived English Learner (EL) students first enrolled in U.S. schools within 12 months from the first day of the previous year’s test window (i.e., students first enrolled after 4/1/2021) are not included in 2021-2022 assessment performance results reporting. Students who have exited EL status within 2 years prior to 2021-2022 assessments are included in performance reporting in their designated group.
The following data sources were used to create this dataset: LEA (Local Education Agency) Information Management System, PARCC and MSAA Results Data, and Demographic, Enrollment, and Assessment Participation Certification.
The dataset is available here.
Load the packages (tidyverse, RColorBrewer, ggfortify, plotly, and GGally).
Set the working directory and load the dataset from a .csv file.
Make column headers lowercase and remove spaces. Then view the data.
## # A tibble: 6 × 18
## leacode leaname schoolcode schoolname subject `testedgrade/subject`
## <chr> <chr> <dbl> <chr> <chr> <chr>
## 1 001 District of Colum… 1058 Bard High… Math Geometry
## 2 001 District of Colum… 1058 Bard High… Math Algebra I
## 3 001 District of Colum… 1058 Bard High… ELA English II
## 4 001 District of Colum… 1058 Bard High… ELA All
## 5 001 District of Colum… 1058 Bard High… ELA All
## 6 001 District of Colum… 1058 Bard High… ELA All
## # ℹ 12 more variables: gradeofenrollment <chr>,
## # totalnumberofvalidparcctesttakers <chr>,
## # `totalnumberofparcctesttakerswithperformancelevel4+` <chr>,
## # `percentoftotalparcctesttakerswithperformancelevel4+` <chr>,
## # numberofstudentswithdisabilitieswhotookparcc <chr>,
## # `numberofstudentswithdisabilitieswhotookparccwithperformancelevel4+` <chr>,
## # `percentofstudentswithdisabilitieswhotookparccwithperformancelevel4+` <chr>, …Clean the values of the columns I am interested in - remove the % symbol and convert “DS” and “n<10” to n/a.
Change the data type for each column to integers and numbers as needed.
Preliminary visualizations
Differences in the total valid number of PARCC test takers for ELA and Math:
The number of valid PARCC math test takers is slightly lower.
Differences in number of alternate assessment takers by LEA:
DC Public Schools (LEA Code 001) and St. Coletta Special Education Public Charter School (LEA Code 143) are the only LEAs with students who take the alternate assessment. None of the students at St. Coletta’s take PARCC since this school specializes in serving students with significant disabilities. Therefore, I will not explore comparisons of PARCC and MSAA in this project.
Percentage of students with 4+ scores on PARCC by LEA:
DCPS has the most schools represented in this dataset out of any other LEA. Many of the other LEAs represent a single school or a small group of schools. Comparing the data by school instead of by LEA will decrease the possibility of misleading visualizations/inaccurate interpretation.
I will sort and filter the data to look at schools who have at least 90 PARCC test takers and at least 10 students with disabilities who took PARCC. To avoid duplicating results, I will look at data for all grade levels (instead of by specific grade bands, since grade-band-level results would also be aggregated in the “all grade levels” row).
## leacode leaname schoolcode
## 1 001 District of Columbia Public Schools 239
## 2 001 District of Columbia Public Schools 246
## 3 001 District of Columbia Public Schools 254
## 4 001 District of Columbia Public Schools 254
## 5 001 District of Columbia Public Schools 254
## 6 001 District of Columbia Public Schools 257
## schoolname subject testedgrade.subject gradeofenrollment
## 1 Garrison Elementary School Math All All
## 2 Hardy Middle School Math 8 All
## 3 Janney Elementary School ELA 3 All
## 4 Janney Elementary School ELA 5 All
## 5 Janney Elementary School Math 5 All
## 6 Ketcham Elementary School Math All All
## totalnumberofvalidparcctesttakers
## 1 94
## 2 90
## 3 97
## 4 92
## 5 92
## 6 99
## totalnumberofparcctesttakerswithperformancelevel4.
## 1 31
## 2 17
## 3 82
## 4 79
## 5 NA
## 6 6
## percentoftotalparcctesttakerswithperformancelevel4.
## 1 32.98
## 2 18.89
## 3 84.54
## 4 85.87
## 5 NA
## 6 6.06
## numberofstudentswithdisabilitieswhotookparcc
## 1 26
## 2 17
## 3 15
## 4 10
## 5 10
## 6 12
## numberofstudentswithdisabilitieswhotookparccwithperformancelevel4.
## 1 3
## 2 NA
## 3 10
## 4 5
## 5 NA
## 6 NA
## percentofstudentswithdisabilitieswhotookparccwithperformancelevel4.
## 1 11.54
## 2 NA
## 3 66.67
## 4 50.00
## 5 NA
## 6 NA
## numberofstudentsw.disabilitiestakingparccw.registeredaccommodations
## 1 NA
## 2 NA
## 3 NA
## 4 NA
## 5 NA
## 6 NA
## numberofstudentsw.disabilitiestakingparccw.registeredaccommodations.performancelevel4.
## 1 NA
## 2 NA
## 3 NA
## 4 NA
## 5 NA
## 6 NA
## numberofalternateassessmenttesttakers
## 1 NA
## 2 NA
## 3 NA
## 4 NA
## 5 NA
## 6 NA
## numberofalternatetesttakerswithperformancelevel3.
## 1 NA
## 2 NA
## 3 NA
## 4 NA
## 5 NA
## 6 NA
## percentofalternatetesttakerswithperformancelevel3.
## 1 NA
## 2 NA
## 3 NA
## 4 NA
## 5 NA
## 6 NAAdd a new column to represent students without disabilities (swod) who took PARCC, achieved performance level 4+, and their percentage.
Visualization including school name, subject (ELA or math), percentage of students who scored 4+, and percentage of students with disabilities who scored 4+.
Data tidying
First, I edited all the column names by removing capitalization and spaces so I could use the column names in my code.
I selected columns of interest to tidy further by removing “%” from percentage values. From these columns of interest, I also replaced “DS” and “n<10” with NA, since “DS” and “n<10” represent information that was not aggregated for the dataset. The documentation for this dataset includes information about how it was determined that these values should replace the raw data that was collected. (However, there is no information about why “DS” is different from “n<10”)
I changed the data type of columns representing quantitative values from character to integer (counts) and numeric (percentages) as needed.
I used the sort and filter functions to create a set of schools with at least 90 students overall who took PARCC and at least 10 students with disabilities who took PARCC. I did so because there were numerous schools with small numbers of students who took PARCC and low numbers of students with disabilities who took PARCC (or did not take any standardized assessments). Given the size of this dataset and disproportionate representation of certain schools, retaining the schools with smaller values would have made it challenging to interpret a visualization generated from it. I chose to look at school-level data since grade-band level data would have been duplicated in the “all grade levels” row. This duplication would have been misleading in a visualization.
I used the mutate function to add the following columns for each school: students without disabilities (SWOD) who took PARCC, SWOD who achieved performance level 4+, and the percentage of SWOD who achieved performance level 4+.
What this visualization represents
The percentage of students scoring 4+ appears to correlate with the percentage of students with disabilities scoring 4+.
In the data included in this visualization, nearly all the students with disabilities (except for one school) took PARCC without any accommodations. However, there were several schools in the original dataset where students with disabilities received accommodations for PARCC. They were not included in this visualization because they did not have at least 10 students with a performance level of 4+. It is unclear why this information was not included because all of the data is already anonymous.
Given how large this dataset is, I was surprised that this visualization showed less information than I was expected. There are a few possible reasons:
This visualization illustrates the dilemma faced by educators who are told to “bridge the gap” between students who attain high scores and students with lower scores. If a student is identified as a student with a disability at school, they are unable to perform on grade level without specialized instruction (which is much more extensive than just testing accommodations), according to IDEA eligibility criteria. This is likely to exclude them from the sample used for this visualization.
The source of this dataset did not include any information about the number of students who had opted out of PARCC or MSAA. Anecdotally, I know that the opt out rate was much higher in 2021-22 compared to previous years. This was likely because the 2021-22 school year was the first full year of in-person learning at DCPS after the COVID-19 pandemic. I would like to know why the opt-out information is not included in the source of this dataset; it does not seem easy to find elsewhere either. We cannot measure and learn about what we are not aware of, so I hope that data like this is not excluded in future collections.
I was also surprised that Janney ES had a large percentage of students with and without disabilities who demonstrated proficiency in ELA, but none in math. (This also occurred with Capital City PCS, Columbia Heights EC, and Wilson HS- now known as Jackson-Reed.) Garrison ES was the opposite, with some students demonstrating proficiency in math but none in ELA.
Further exploration
It was difficult to work with this dataset because of how the data was formatted, the lack of transparency around why certain information was not included, and the lack of contextual information (e.g., number of students who opted out of PARCC, total enrollment, etc.)
- I was unable to assign 2 different colors for categorical variables with the data I was able to sort and filter. I was unable to keep the gradient for percentage of students intact while distinguishing color for each school or subject.
- I tried to separate the stacked bars for Janney ES (one represents 3rd grade and the other represents 5th grade) but was unable to do so without removing the entire school or adding several blank spaces to the visualization for other schools, and was not able to add labels to these categories. The current visualization is confusing but changing it within the timeframe for this assignment would have made it even harder to interpret.
- If I had more time, I would have liked to sort performance by grade level across schools.
- If I had even more time, I would have liked to map schools by ward/neighborhood and see if PARCC scores are related to public/charter status, median income by neighborhood, Title 1 status, tree cover, proximity to healthcare facilities, proximity to libraries, crime rate, and other variables that are linked to SES (and thus affect educational outcomes) but may not be mentioned often.
- Look at tested grade/subject instead of subject- this is not the best way of displaying this information.
I created a new column to separate students with/without disabilities in the dataset but found it was not necessary and might create confusion.
I would have liked to arrange the data by elementary/middle/high school.
I wanted to pivot the new dataframe so the percentage of students with and without disabilities with scores >4 would be in the same column for the visualization. This probably would have made it easier to assign different colors to categorical variables. Code that I started:
#```{r} parcc90c |> pivot_longer( cols = pctswodperformancelevel4:percentofstudentswithdisabilitieswhotookparccwithperformancelevel4., names_to = “pctlevel4”, values_to = “value” )