Category Archives: Item Analysis

Areas of Celebration and Exploration

After a brief interlude, it’s time to get back to the blog series I started recently about analyzing assessments.

  • In the first post, I shared the importance of digging into the questions, not just the standards they’re correlated to.
  • In the second post, I talked about how understanding how a test is designed can help us better understand the results we get.
  • In the third post, I shared how I learned to organize assessment data by item difficulty and the implications for supporting our students.
  • In this post, I’d like to talk about another way to look at assessment data to uncover areas of celebration and areas of exploration.

Let’s get started!


In my previous post I shared the order of questions based on item difficulty for the 2018 5th grade STAAR for the entire state of Texas. Here it is again:

2018-G5-Item-Difficulty-Sort

According to this ordering, question 9 was the most difficult item on the test, followed by question 18, question 8, and so on down to question 10 as the least difficult item (tied with questions 2 and 4).

Here’s my question: What is the likelihood that any given campus across the state would have the exact same order if they analyzed the item difficulty just for their students?

Hopefully you’re like me and you’re thinking, “Not very likely.” Let’s check to see. Here’s the item difficulty of the state of Texas compared to the item difficulty at just one campus with about 80 students. What do you notice? What do you wonder?

2018-G5-Texas-vs-Campus

Some of my noticings:

  • Questions 8, 9, 18, and 21 were some of the most difficult items for both the state and for this particular campus.
  • Question 5 was not particular difficulty for the state of Texas as a whole (it’s about midway down the list), but it was surprisingly difficult for this particular campus.
  • Question 22 was one of the most difficult items for the state of Texas as a whole, but it was not particularly difficult for this campus (it’s almost halfway down the list).
  • Questions 1, 2, 10, 25, and 36 were some of the least difficult items for both the state and for this particular campus.
  • Question 4 was tied with questions 2 and 10 for being the least difficult item for the state, but for this particular campus it didn’t crack the top 5 list of least difficult items.
  • There were more questions tied for being the most difficult items for the state and more questions tied for being the least difficult items for this particular campus.

My takeaway?

What is difficult for the state as a whole might not be difficult for the students at a particular school. Likewise, what is not very difficult for the state as a whole might have been more difficult than expected for the students at a particular school.

But is there an easier way to identify these differences than looking at an item on one list and then hunting it down on the second list? There is!

This image shows the item difficult rank for each question for Texas and for the campus. The final column shows the difference between these rankings.

2018-G5-Rank-Order

 

Just in case you’re having trouble making sense of it, let’s just look at question 9.

2018-G5-Rank-Order-Q9

As you can see, this was the number 1 most difficult item for the state of Texas, but it was number 3 on the same list for this campus. As a result, the rank difference is 2 because this question was 2 questions less difficult for the campus. However that’s a pretty small difference, which I interpret to mean that this question was generally about as difficult for this campus as it was for the state as a whole. What I’m curious about and interested in finding are the notable differences.

Let’s look at another example, question 5.

2018-G5-Rank-Order-Q5

This is interesting! This question was number 18 in the item difficulty for Texas, where 1 is the most difficult and 36 is the least difficult. However, this same question was number 5 in the list of questions for the campus. The rank difference is -13 because this questions was 13 questions more difficult for the campus. That’s a huge difference! I call questions like this areas of exploration. These questions are worth exploring because they buck the trend. If instruction at the campus were like the rest of Texas, this question should have been just as difficult for the campus than for the rest of the state…but it wasn’t. That’s a big red flag that I want to start digging to uncover why this question was so much more difficult. There are lots of reasons this could be the case, such as:

  • It includes a model the teachers never introduced their students to.
  • Teacher(s) at the campus didn’t know how to teach this particular concept well.
  • The question included terminology the students hadn’t been exposed to.
  • Teacher(s) at the campus skipped this content for one reason or another, or they quickly glossed over it.

In case you’re curious, here’s question 5 so you can see for yourself. Since you weren’t at the school that got this data, your guesses are even more hypothetical than there’s, but it is interesting to wonder.

2018-G5-Q5

Let me be clear. Exploring this question isn’t about placing blame. It’s about uncovering, learning what can be learned, and making a plan for future instruction so students at this campus hopefully don’t find questions like this so difficult in the future.

Let’s look at one more question from the rank order list, question 22.

2018-G5-Rank-Order-Q7

This is sort of the reverse of the previous question. Question 7 was much more difficult for the state as a whole than it was for this campus. So much so that it was 7 questions less difficult for this campus than it was for the state. Whereas question 5 is an area of exploration, I consider question 7 an area of celebration! Something going on at that campus made it so that this particular question was a lot less difficult for the students there.

  • Maybe the teachers taught that unit really well and student understanding was solid.
  • Maybe the students had encountered some problems very similar to question 7.
  • Maybe students were very familiar with the context of the problem.
  • Maybe the teachers were especially comfortable with the content from this question.

Again, in case you’re curious, here’s question 22 to get you wondering.

2018-G5-Q22

 

In Texas this is called a griddable question. Rather than being multiple choice, students have to grid their answer like this on their answer sheet:

2018-G5-Q22-Grid

Griddable items are usually some of the most difficult items on STAAR because of their demand for accuracy. That makes it even more interesting that this item was less difficult at this particular campus.

We can never know exactly why a question was significantly more or less difficult at a particular campus, but analyzing and comparing the rank orders of item difficulty does bring to the surface unexpected, and sometimes tantalizing, differences that are well worth exploring and celebrating.

Just this week I met with teams at a campus in my district to go over their own campus rank order data compared to our district data. They very quickly generated thoughtful hypotheses about why certain questions were more difficult and others were less so based on their memories of last year’s instruction. In meeting with their 5th grade team, for example, we were surprised to find that many of the questions that were much more difficult for their students involved incorrect answers that were most likely caused by calculation errors, especially if decimals were involved. That was very eye opening and got us brainstorming ideas of what we can work on together this year.


This post wraps up my series on analyzing assessment data. I might follow up with some posts specifically about the 2018 STAAR for grades 3-5 to share my analysis of questions from those assessments. At this point, however, I’ve shared the big lessons I’ve learned about how to look at assessments in new ways, particularly with regards to test design and item difficulty.

Before I go, I owe a big thank you to Dr. David Osman, Director of Research and Evaluation at Round Rock ISD, for his help and support with this work. And I also want to thank you for reading. I hope you’ve come away with some new ideas you can try in your own work!

Difficult

This post is the third in a series where I’m sharing how I’ve changed the ways that I look at assessments and assessment data.

  • In the first post, I shared the importance of digging into the questions, not just the standards they’re correlated to.
  • In the second post, I talked about how understanding how a test is designed can help us better understand the results we get.
  • In this post, I’d like to share one of the ways I’ve learned how to analyze assessment results.

Let’s get started!


Do you know what the most difficult item on an assessment is?

  • Is it the one with a pictograph with a scaled interval that involves combining the values from several categories?
  • Is it the multi-step story problem involving addition, subtraction, and multiplication?
  • Is it the one about matching a set of disorganized data with the correct dot plot out of four possible answer choices?

Here’s the thing I learned from Dr. Kevin Barlow, Executive Director of Research and Accountability in Arlington ISD, no matter how much time and effort someone spends designing an item, from crafting the wording to choosing just the right numbers, the only way to determine the difficulty of an item is to put it in front of students on an assessment. After students are finished, take a look at the results and find the question where the most students were incorrect.

You found it! That’s the most difficult item on the assessment.

Through their responses, our students will tell us every single time which question(s) were the most difficult for them. It’s our responsibility to analyze those questions to determine what made them so challenging.

Fortunately, the Texas Education Agency provides this information to us in Statewide Item Analysis Reports. Unfortunately, it starts out looking like this:

2018-G5-Item-Analysis-TEA

This is a great first step, but it’s not terribly useful in this format. You can’t glance at it and pick out anything meaningful. However, if I copy this data into a spreadsheet and sort it, it becomes so much more useful and meaningful:

2018-G5-Item-Difficulty-Sort

Now I’ve sorted the questions based on how students performed, from the item most students answered incorrectly (#9 was the most difficult item on this test) to the item the least number of students answered incorrectly (#2, #4, and #10 were tied for being the least difficult items on this test). It’s interesting to think that #9 and #10, back to back, turned out to be the least and most difficult for 5th graders across the state of Texas!

The items highlighted in red were the most difficult items for 5th graders. Remember, it doesn’t matter how the questions were designed. These items were the most difficult because the least number of students answered them correctly.

The items highlighted in blue, on the other hand, were the least difficult items for 5th graders in Texas. I’m intentional about calling them the least difficult items. We might be inclined to call them the easiest items, but that obscures the fact that these questions were still difficult enough that 14-17% of all Texas 5th graders answered them incorrectly. To put some real numbers with that, anywhere from 56,000 to 68,000 students answered these “easy” items incorrectly. These items were clearly difficult for these students, but they were the least difficult for the population of 5th graders as a whole.

Now what?

We might be inclined to go to the items in red and start analyzing those first. Great idea! But for whom?

Well, since they were the most difficult items, meaning the most students missed them, we should use these items to teach all of our students, right? Clearly everyone had issues with them!

I’m going to disagree with that.

These items were difficult even for some of our strongest students. If they struggled, then the last thing I want to do is bring this level of challenge to all of my students, especially those who struggled throughout the test. Rather, I’ll analyze the most difficult items to get ideas to provide challenge to my higher performing students. These kinds of questions are clearly structured in a way that gets them thinking, challenges them, and perhaps even confuses them. That’s good information to know!

(Please don’t misinterpret this as me saying that I don’t want to challenge all students. Rather, I want to ensure all students are appropriately challenged, and that’s what I’m trying to identify through this kind of analysis. Read on to see what I mean.)

But what about students who struggled throughout the test? For those students, I’m going to analyze the least difficult items. In this case, 14-17% of students in Texas answered even these items incorrectly. These items posed a challenge for quite a number of students, and I want to analyze the items to figure out what made them challenging for these students.

Let’s pretend that this is school data instead of Texas data, and let’s pretend we’re a team of 6th grade teachers analyzing 5th grade data for our 200 6th graders. That would mean at least 28-34 students in our 6th grade did not do well on these least difficult items when they took 5th grade STAAR last spring. That’s a pretty significant number of kids! They could for sure benefit from some form of intervention based on what we learn from analyzing these items.

And that’s where I’m going to leave this in your hands! Here is a document where I’ve collected the most difficult and least difficult items from the 2018 5th grade STAAR. These are the actual test questions along with the percentage of students who selected each answer choice. Spend a little time analyzing them. Here are some questions to guide you:

  • What are the features of each question? (How is the question constructed? What are its components and how are they put together in the question?)
  • Why do you suppose the features of a given question made it more/less difficult for students?
  • What mathematical knowledge and skills are required to be successful with each question?
  • What non-mathematical knowledge and skills are required to be successful with each question?
  • What can you learn from analyzing the distractors? What do they tell you about the kinds of mistakes students made or the misunderstandings they might have had?
  • What lessons can we learn from these questions to guide us in how we support our students? (We don’t want to teach our students these exact questions. That’s not terribly useful since they won’t be taking this exact test again. Rather, seek out general themes or trends that you observe in the questions that can guide your classroom instruction and/or intervention.)

I’ve opened up the document so that anyone can comment. If you’d like to share your thoughts on any of the questions, please do! I look forward to reading your thoughts about the least and most difficult items on the 2018 5th grade STAAR.

I’m giving you a very small set of questions to analyze right now. You may or may not be able to generalize much from them depending on your own experiences analyzing assessment items. However, it’s worth doing regardless of your experience, because now the repertoire of items you’ve analyzed will be that much larger.

As for myself, I’ve been analyzing assessment items like this for several years. What I’d like to do in my next post is share some of the lessons I’ve learned from this analysis across multiple years. I do feel like there are consistent trends (and a few surprises) that can inform our work in ways that simultaneously align with high-quality math instruction (because ultimately this is what I care much more about than testing) while also ensuring students are given the supports they need to succeed on mandatory high stakes tests (because they are a fact of life and it’s our responsibility to ensure students, especially those who are relying on school for this support, are prepared for them).

 

 

Misplaced Priorities

Every spring thousands upon thousands of Texas students take the State of Texas Assessments of Academic Readiness (STAAR for short). It’s a one-day snapshot meant to evaluate a year of learning within a subject area. Even though many disagree with one-time events as assessments of learning, the fact of the matter is that they are a reality for us and our students. Because these assessments carry so much weight, we pore over the data they generate, often looking for standards where our students performed poorly so we can identify what to focus on in our instruction and intervention.

But what if I told you this well-intentioned practice may be sending us in unproductive directions? Rather than focusing on what our students really need, we may be spending time on topics and/or skills that are not the priority.

Let me illustrate what I mean with a story. I was working with a 4th grade team after a district benchmark we call STAAR Ready. Every spring in my district we give our students a released STAAR to gauge readiness for the actual STAAR coming up in May. Afterward, teams analyze the data to determine which topics to revisit and which students to put into intervention groups.

As I met with this 4th grade team, they showed me a list of the low-performing TEKS (Side note: this is what we call our standards in Texas – the Texas Essential Knowledge and Skills, TEKS for short) they had identified after analyzing the STAAR Ready data. One of the TEKS jumped out at me immediately because I was familiar with the test:

TEKS 4.4A add and subtract whole numbers and decimals to the hundredths place using the standard algorithm;

I asked them to tell me more, and the team told me they had identified students who performed poorly on the questions correlated to this standard. They created an intervention group with these students to work on adding and subtracting whole numbers and decimals to make sure they could do these computations accurately.

I followed up with a question, “Have you looked at the actual questions correlated to these TEKS?” Because they were looking at so much data and so many standards, they hadn’t gotten back into the test. Instead they’d just been identifying high-priority TEKS based on student performance on the questions.

I pulled up the test and showed them this question that had immediately come to mind when they told me they were making a group focused on TEKS 4.4A:

TEA-STAAR-4-2016-Item-34

Source: Texas Education Agency, STAAR Math, Grade 4, Item 34

Take a moment and analyze the question.

  • Can you see how it involves adding and/or subtracting with whole numbers and/or decimals?
  • But what other skills are involved in answering this question correctly?
  • What features of the problem might have made it more difficult for the students to answer correctly?

As it turns out, this was an incredibly difficult problem for students! When it was given to students on the actual STAAR in spring 2016, only 43% of students across the state of Texas were able to answer correctly. That means 57% of Texas 4th graders, or roughly 209,390 students, couldn’t find the total cost of three items in a shopping basket. That’s…concerning.

In my own school district, we used the 2016 released STAAR as our STAAR Ready in spring 2017. This allowed me to collect data Texas doesn’t make available to everyone. When we gave the test in spring 2017, the problem was nearly as difficult for our students. About 48% of students in my district answered it correctly. I was also able to determine this was the 6th most difficult item on the entire test of 48 questions!

What’s going on? A lot actually, for such a short question. For starters, key information is spread across two sentences. The first sentence of the problem indicates the quantities of items purchased – 1 hat and 2 skirts. The second sentence indicates their prices. This is subtle, but separating that information across two sentences upped the level of difficulty significantly for 9 and 10 year olds. Students who are not reading closely can quickly jump to the conclusion that they only need to add the two prices shown without realizing that one of those prices needs to be used twice.

The second feature of this problem that ups the difficulty is the fact that it is an open response question, not multiple choice. On this kind of question, a student’s answer has to be absolutely 100% accurate. If they’re off by even 1 penny, the answer is marked wrong. No pressure, kids!

I was curious which feature made the problem more difficult for the students in my district, so I dove into the data. One thing I had available that Texas doesn’t release is the actual answers every student submitted for this problem. I was able to analyze roughly 3,600 answers to see what students were doing. Here’s what I found out.

While only 48% of students got this question correct, there was a chunk of students whose answers were in the ballpark. These are kids who likely made a small calculation error. Unfortunately, if I calculate the percent of students who got it right or reasonably close, that only brings it up to 51% of our 4th graders. That’s not terribly impressive.

So what was everyone else doing? Here’s where it gets interesting. I predicted that these students only found the cost of 1 hat and 1 skirt, and it turns out that’s exactly what 33% of students in my district did. Nearly 1,200 students failed to comprehend that the total cost is composed of a hat, a skirt, and another skirt.

Going back to the team I was working with, I asked, “So now that we’ve analyzed this question, do you think the issue is that your students are struggling with adding and subtracting whole numbers and decimals?” We talked about it and they agreed that the bigger issue is how their students read and comprehend word problems.

Looking just at the standards is a very limiting view of analyzing data. There are often many different ways to assess a standard, and if we don’t take the time to look at the exact questions our students interact with, we might be missing critical information. Had this team done an intervention on pure addition and subtraction of whole numbers and decimals their kids would have gotten better at those skills for sure. But is that really what they needed?

Over the past year, I’ve been analyzing assessment data differently than in the past. In follow up posts I’d like to share some of that with you. In the meantime, please dive into your assessments and analyze those questions, not just the standards. You’ll hopefully come away with a truer picture of what’s challenging your students so that you can more accurately target with what and how to support them.