by Dave Kinnaman and LouAnn Ballew
Copyright 1999, Dave Kinnaman, all rights reserved.
Although there are reports of questions covering unspecified new material, the full breadth of the question sets may have been previously unknown or unreported.
Some new interactive questions may require you to place values into boxes onscreen, such as filling in a DHCP scope. You may be required to mark IP addresses by dragging text labels of "valid" "or invalid" to the IP address. Another interactive method is one where you must drag text labels to match column A with column B. As with other Microsoft tests, any wrong answer is deadly - no "parial credit" is allowed. A single wrong pair causes you to lose the whole question, even though you may correctly match other parts of the question.
New questions can be quietly introduced in an adaptive test just as easily as they could be on the "old fashioned" exams. As yet, however, the official Microsoft Preparation Guidelines have remained unchanged.
Efficiently finding the candidate's level of knowledge is the supreme goal of testing now. No feedback is provided other than your overall score and whether or not you passed.
Previously, both the real tests and the PEP tests provided feedback on each tested category in the official Microsoft Preparation Guidelines. The new adaptive tests are only efficient at doing one thing - finding out if you pass or not. These new adaptive tests are simply not intended to help you better prepare yourself.
It once was possible to gauge your understanding against another MCP by comparing your scores. Because all candidates received the same or similar forms of the exam, and each candidate received all the questions on that exam, higher scores clearly meant better understanding. It is no longer clear that these comparisions are valid - the test may end when you have achieved a passing score, rather than when you have achieved the highest score you could have achieved under the old system.
Questions were grouped - probably by the official Microsoft Preparation Guide categories. The difficulty of every question is known because it has been recorded while thousands of us took these tests. So, each question has a known pass rate, and a known Preparation Guide category.
From the scores achieved by various certificate holders and other demographics, all questions can be rated for how predictive they are of one or more arbitarily selected indicator(s) of success.
Combining the information about how predictive of success a question is with the previously calculated data about question difficulty allows psychometricians to choose the most predictive questions that are either harder or easier within the same test category.
In this way, a smaller group of highly predictive questions on each category or area of the test can be selected. Easy questions have small point values, and harder questions are given larger point values.
Microsoft adaptive tests start with questions of easy-to-moderate difficulty. Correct answers lead to questions of increasing difficulty, incorrect answers lead to easier questions.
This process continues only until the adaptive test determines the candidate's ability. This means that during a test, even though a candidate might well be able to answer many more questions correctly, the test ends when the candidate has surely passed or surely failed. As soon as the test knows that you could not possibly pass, or that you have certianly already passed, testing ends.
|Same: Advanced Warning From Microsoft|
By David Foster of Galton Technologies, Inc., and Tina Koyama
Created: Monday, January 26, 1998 2:24:00 PM (88kb file download)
Computer Adaptive Testing Sample
Last Modified: Thursday, 01-Oct-98 18:26:56 GMT (zipped 333kb file)
Frequently Asked Questions About Adaptive Testing
Last Modified: Wednesday, 30-Sep-98 22:13:22 GMT
Overview of New Test Formats
Last Modified: Friday, 25-Sep-98 01:39:56 GMT
Based on the official Microsoft Preparation Guide, and on the free PEP practice tests, candidates are able to gauge the progress of their studies by improved scores on Planning, Installation and Configuration, Monitoring and Optimization, Troubleshooting, and so on. Real exams used to provide, in addition to a total overall exam score, partial scores on the categories appropriate for each exam.
Partial scores, or sub-scores, are not available for adaptive tests. Partial scores may be withdrawn from other exams, as well.
The carefully organized question set for each adaptive test is now presented in a way that is designed to reveal the candidate's level of ability with a bare minimum number of questions. During your adaptive test, if you receive too many questions, things may be going poorly. These new tests may be designed to test the usual candidate with as little as 35 or fewer questions. Candidates are reporting successful 15 minute tests! Failure may take slightly longer.
The non-disclosure agreement (NDA) and the previously mandatory demographic survey have been shifted from the beginning of the exam to the end. The demographic survey is mandatory at most testing centers.
On an adaptive test, you cannot go back and change your answer on a previous question. You cannot skip a question. These changes make the Microsoft tests more stressful, even if these adaptive tests do measure more efficiently.
Be ABSOLUTELY finished with each question before moving on to the next item in an adaptive test.
For many months MCPs have been aware that Microsoft is re-thinking test security. First we heard that third re-tests were being required to wait two weeks. Then the mandatory Non-Disclosure Agreement was introduced.
Adaptive testing further tightens the security around the actual test questions. An absolute minimum of questions will be exposed to any one candidate - making the job of Transcender writers both tougher and more costly. Ironically, adaptive tests actually make any correct intelligence available from braindumps even more valuable!
Beta examinees were once allowed to know their scores within 6 to 8 weeks after taking a Beta test. Now, they are not told their scores until the actual test is released into duty -- possibly much longer than the previous wait.
By preventing candidates from knowing their partial scores, and by delaying beta-test takers from getting their results, Microsoft appears to be hoping that they can reduce the information flow around the test questions. It is logical that braindump Internet sites may be the next target for the finger-in-the-dike team.