Thursday, June 18, 2009

Care must be taken when administering benchmark assessments to subsets of students or to students from multiple grade levels

Galileo K-12 Online benchmark assessments serve two functions simultaneously. One is to provide teachers with timely feedback regarding which standards their students have and have not mastered. The other is to forecast the students’ likely performance on the high stakes statewide assessment such as AIMS in Arizona or MCAS in Massachusetts. Both of these functions are equally important, and in most cases both goals are achieved in harmony by the single benchmark assessment. However, there are some cases where the two goals are in conflict. In today’s post, I want to alert district administrators to a potential problem and to give them a way to avoid it when planning benchmark assessments.

In the typical scenario, a benchmark assessment is given to all students in the district in a given grade level. For example, all fifth-graders in the district might take a fifth-grade math benchmark assessment. It is expected that all of these students will also take the fifth-grade math high-stakes statewide assessment. This is important because the benchmark assessment must be aligned to the statewide assessment in order to generate cut scores for performance levels and to forecast student performance on the statewide assessment. If the same set of students is expected to take both the benchmark assessment and the statewide assessment, then the comparison between the two assessments is essentially a comparison of apples to apples, and all is well. The cut scores that are calculated for the benchmark assessment should provide accurate forecasts of student performance on the statewide assessment and, in fact, the accuracy rate for Galileo K-12 Online benchmark assessments is quite high (see the Galileo K-12 Online Technical Manual.)

There are cases, however, where the set of students taking a benchmark assessment is not the same as the set that will be taking the statewide assessment. In these cases, the calculation of accurate cut scores for benchmark assessments becomes more complicated. A common scenario is one in which advanced 8th-graders are taking a high school algebra course and, quite reasonably, they take the high school math benchmark assessments instead of the 8th grade math benchmark assessments. This makes perfect sense for the first goal of benchmark assessments: providing feedback to teachers regarding student mastery of state standards. It does, however, create problems for the goal of forecasting student performance on the statewide assessment. In most cases these students will be taking the 8th grade statewide assessment, and not the high school statewide assessment, and so the comparison when calculating cut scores becomes one of apples to oranges.

In order to calculate accurate cut scores for the high school math benchmark assessment in the above scenario, the scores from the 8th grade students must be removed from the data set, so that the set of students on the benchmark assessment will be the same as the set of students who will be taking the high school statewide assessment. Additionally, care must be taken when calculating the cut scores for the 8th grade math benchmark assessment. This is because a specific region of the student distribution, the advanced students, will not be present in the distribution of scores for the 8th grade benchmark. If no adjustment is made to account for the absence of the advanced students, then the cut scores that are calculated will be too low, and too many students will be classified as being likely to pass the statewide assessment. This, of course, will result in rude surprises when the statewide assessment results come in.

The take-home message, then, is to be sure to be clear about who will be taking benchmark assessments when you are planning them. Steps can be taken in cases such as the one described here to make sure that the cut scores on benchmark assessments are accurate, but only if ATI knows about the unusual circumstances in advance. If you are designing benchmark assessments in Galileo K-12 Online and there will be any out-of-grade testing, or if the set of students on the benchmark assessment will not be the same as the set that is taking a particular statewide assessment, please let your Field Services or Educational Management Services representative know right away. Forearmed with as much information as possible, ATI can work with your district to make sure that the benchmark assessments provide accurate forecasts of student performance on statewide assessments as well as providing timely feedback regarding the mastery of standards to classroom teachers.

No comments: