Officials say test errors may never be sorted out

ATLANTA - The scoring errors on a standardized test used by schools to measure student performance in basic skills may never be fixed properly, state officials said Thursday. <br> <br> ``We don&#39;t know what the problem is,&#39;&#39; said Mike Harmon, program manager for test administration at the Georgia Department of Education. <br> <br> ``It is incredibly frustrating. And even more frustrating than it is for us here,&#39;&#39; Harmon said, ``is the frustration for teachers and for parents and for children. <br> <br> ``They worked so hard and they got these results back and they expected them to be an accurate representation, and we don&#39;t believe that it is,&#39;&#39; he said. <br> <br> The state requires the Stanford 9 test be given each spring to public school students in grades 3, 5 and 8 to measure their performance in math and reading. <br> <br> School administrators say the test helps determine placement for students, deciding whether to put them in advanced or remedial classes. Some districts use the Stanford 9 to determine bonus money for administrators, teachers and staff. <br> <br> The Stanford 9 ``is not just another test,&#39;&#39; said Martha Greenway, executive director of planning, research and policy with Fulton County schools. ``It&#39;s a critical measure of how our students are doing. And we use that to plan for the future.&#39;&#39; <br> <br> Some school officials have downplayed the impact, saying they have other ways to measure students&#39; abilities, such as the state&#39;s curriculum test and the judgment of teachers and principals. <br> <br> But the testing company that developed the Stanford 9, Harcourt Educational Management, has brought in three testing experts to sort out the mess, at Harcourt&#39;s expense. The state paid $690,000 for the tests. <br> <br> State officials first noticed there was a problem with this year&#39;s scores in May. In about one-third of the subject areas, scores went up or down by large margins compared with the year before, Harmon said. A change of 1 to 3 percentage points is all that should be reasonably expected for a statewide average, he added. <br> <br> Harcourt checked the tests and found they were scored accurately. No correct answers were marked as incorrect, or vice versa. The problem appears to be in ``equating&#39;&#39; the 2002 test with the 2001 test, making sure they are of the same difficulty level so that scores on the two tests can be compared. <br> <br> Harcourt spokesman Richard Blake said the company used a commonly accepted equating method, ``but in this case it produced anomalous results.&#39;&#39; <br> <br> ``We are working as hard as we can to get this resolved as quickly as we can,&#39;&#39; Blake said. ``We are sympathetic that these problems, these delays, have caused problems for people.&#39;&#39; <br> <br> In June, the state Department of Education decided to make the test optional for districts next school year.

http://accesswdun.com/article/2002/7/192322