How can teachers use Tableau to identify students who need extra attention?

Posted On
Posted By Jonathan Rauh

The kinds of data teachers use tends to fall into two different categories, achievement data and non-academic data. Achievement data may be further broken down based on how the data are collected. For example test scores present one type of performance data and in-class performance presents a different type of data. Whether a teacher is using pre/post or interim assessments, or is focused on in-class performance such as tests, quizzes, and homework, their analysis tends to focus on descriptive statistics of student performance including, highest and lowest (min/max), average performance and median performance. Unfortunately these do not provide teachers with much insight on how to tailor their instruction and provides limited information on which students should be targeted. Using Tableau in the analysis of their data can go a long way to providing teachers with more robust information for decision making.

To begin analyzing their data teachers should start with a question – something that bothers them or something that they want to achieve. For example, a third grade teacher may think, “my students seem to struggle with literal and non-literal meanings of words.” Is this really true? They should then think through the data that can help answer their question.

In the case of a student’s struggle with literal and non-literal meaning we care about “ability” which the teacher may assess on quizzes and tests throughout the year. This is not just the number right or wrong, it also considers the difficulty of the individual question. Instead of looking just at the number of correct responses, the teachers should also look at how difficult the questions were that the students got right versus those they got wrong.

Collect data – but the right data.

Teachers often want to assess the difficulty of their questions to see where students did well and where they struggled. This is fine and it can give good information on what items students in general are struggling with and which questions may need to be dropped. The traditional way of doing this is by determining the difficulty of the question and then determining whether the question was able to discriminate between the high and low performing students. This is traditional Item Assessment and Difficulty Index that teachers learn in their teacher education programs. For a refresher I have provided a link on how to construct these. Assessing these questions really only gives you a post hoc sense of how students are performing and whether or not you have a good question*. How do we know what students to target and whether or not they are really performing better or worse, or need attention.

Collect data on things that influence performance outside of your teaching.

We know that students in poverty tend to score lower than students who are not in poverty. Additionally, we know that students who have special needs or need accommodations tend to perform worse than students who do not have the same challenges. It makes sense therefore to consider these factors when looking at student performance.  Below I have provided a summary of student scores based on a number of factors: whether they scored above the median on their previous year’s ELA assessment, their scores based on their poverty status, and whether the student has an IEP. You will note that I have provided the Average score and the Standard Deviation. The standard deviation is the average dispersion within a data set, e.g. the average distance of any point from the average (arithmetic mean). This is a statistic that far too few teacher use but they really should.

Think about this way, if we took the average salary of every adult in a school then we will probably get a pretty accurate estimate salaries overall. But if Bill Gates happened to be touring the school that day, then on average everyone in the school is a multimillionaire. So averages can be misleading. We need to use the standard deviation to see who is the outlier and who is representative of the norm.

Returning to the example at hand though, if we find that the average score for all students is, for example, 80 and the standard deviation is 6, we can say that a student is scoring 2 standard deviations above the average is likely performing better than the average student. But this also gives us an idea of their performance relative to others in different categories.

Take the Poverty Status of a student for example. If the average performance for students in poverty is 75 with a standard deviation of 5 and we find student in poverty scoring an 83, they would be seen as performing significantly better than their peers who are in poverty but are about the same as the average student overall. Knowing the detrimental effects of poverty on student performance, there may be something about this student that leads them to perform better – a supportive home structure, some technique the teacher is using, etc. This does not tell us why the student is performing better than his peer, only that he is.

How do we implement this in Tableau?

I will be working with a mock dataset that has data on student performance and includes a student ID as well as test scores for each student. We need to begin getting the overall average and standard deviation and then compare that to individual student performance.

First we need to get the average for scores overall. To do this we use a fixed level of detail expression, fixing the overall average to a given subject.

Next, we need to get the overall standard deviation.

Now that we have these two pieces of information, we can see how the average scores of individual students compare to performance overall. To do this, we will used a fixed level of detail expression to determine if the performance of student in a given subject is above the norm (greater than the overall average + 2 standard deviations), below the norm, or in the middle, i.e., not distinguishable from the average.

We now drag our Student Performance Category variable to Columns, drag Student ID to Rows, and drag their scores to rows as well.

We can now see students’ individual scores and whether they are Above the Norm, Below the Norm, or Representative of the Norm. From here we can start to identify patterns among students based on non-academic elements such as race/ethnicity or poverty status. So, for example, we see that White students in poverty perform lower than other students in poverty. And, we can also determine the distance from the norm for any given student.

By simply adding an additional piece of information to our analysis and visualizing it in Tableau, we are able to derive much deeper understanding of student performance, determine which students are in need of additional assistance, and if factors outside the classroom play a role in student performance relative to our classroom overall.

Jonathan received his PhD in Public Policy with an emphasis in Quantitative Analytics focusing in Education and Health Policy. He has held multiple roles in education analytics including Research Director for eLearning for the SC Department of Education, Evaluator for multiple NSF Education Grants, and is the author of multiple articles on eLearning and the use of social analytics in K12. He has also served on the faculty at East Carolina University, College of Charleston, and Virginia Commonwealth University where he taught courses in Statistics, Business Intelligence, and Educational Policy Analysis. Jonathan is currently a Lead Solution Engineer with Tableau at Salesforce.

Related Post

leave a Comment