<p>Okay, I got 37 percentile and a Stanine score of 4 for my Math Achievement section. I got 74 on Quantitive and 90 on Verbal, as well as 64 on Reading (Which was one of my strongest parts on practice tests!). Do you think I'll be penalized for my low score on Math achievement? And the score for Reading is just really weird because I usually get really good in Comprehension.</p>
<p>The ISEE exam was incredibly challenging this year, given the drastic changes made to it in fall of 2009. As with last year, the scores were really off and low (e.g. many of the scores came back in the 5’s in all areas). The reading comprehension part was particularly difficult this year, as the essays/stories were extremely long and difficult to speed read. Therefore, the scale is off, because the percentile was based on scores in the last 5 years (before the drastic changes and when the exam was far less difficult). I believe that all the schools are aware of the drastic changes in the exam and the discrepancies in the results. This is the reason the ISEE is only a piece of the puzzle and considered along with recommendations, your interview, your school record, etc. Good luck!</p>
1 Like
<p>Thanks! That’s reassuring MiltonMomof2 :D</p>
<p>MMof 2: thank you for the explanation; would have been nice if ISEE would have told the parents that don’t you think???</p>
<p>@Rodney, I know right. Most kids were studying from guides for 2009, because no new guides were published that were in accordance to the new standards that ISEE had set. I’m sure schools will take that into account though. Since they made SUCH drastic changes, they should have at least published a guide to go along with the changes.</p>
<p>A small correction-the ISEE is normed on a 3 year pool, not 5.</p>
<p>Thanks for the correction, Sue. She’s absolutely correct - the scale is based on the last 3 years. Apologies for the mistake. </p>
<p>After receiving scores in December/January, several parents and I contacted various educational consultants in both Boston and NYC for an explanation for the craziness of the scoring - particularly in the sections where we expected our children to excel. All of us had tutors for our children, but to no avail. All the consultants told us that it was extremely difficult to prepare for the ISEE this year, because additional changes were made unannounced. Therefore, tutors had a difficult time preparing test takers. After talking to many parents of the kids (all of whom are high achievers and excellent test takers) in my son’s school who applying to ISL schools, it seems that everyone’s scores were unexpectedly low.</p>
<p>I am pretty sure ISEE scores are low across for both the lower and mid/upper levels. For the lower level ISEE in Reading, they give 25 questions but only grade 20 of them and the percentiles are based on the number right out of the 20 they actually scored. Frightening.</p>
<p>What do folks consider low? Which sections were the surprises. Have a S in 8th grade in a school that ends at 8th so whole class takes ISEE and haven’t heard too many surprises–certainly plenty of 7s, 8s and 9s still.</p>
<p>Statistically there are fewer 7’s, 8’s and 9’s this year than there were in test years 2008 and 2009. The prior three years are used to set percentiles, and two of those three years ('08 and '09) were prior to the re-norming that occurred in 2010. I am surprised that people shared their scores with you so openly. But I guess if asked directly by a parent, I would probably say 7, 8 and 9 as well, when in fact there might really be some 5’s in there.</p>
<p>I’m with Belkin39 100%. Low meaning 4s, 5s, and 6s. </p>
<p>I am dubious of anyone who boasts high numbers of any kind (this includes not only test scores of all types, but salary, price tags, etc.). My son had a classmate who had bragged that he thought that the exam was easy and that he got all 9s on the middle level exam. My son felt horrible until this child dropped his ISEE report in the school stairwell and everyone then discovered that the child actually had scores of 4s and 5s. I felt bad for this child, but not too bad, as his lies made many of his classmates feel awful about themselves. I am sure there are kids out there who did in fact score 9s in all four sections of the ISEE, but they are extremely rare for the 2010 exam.</p>
<p>All 9’s are rare in any year, MiltonMom. 45,000 take the test every year, meaning only 450 score in the 99th each year, and that’s across all grades, both sexes, and all geographical regions. It takes far less than all 99’s to score in the 99th, so you can imagine that all 99’s are truly rare. I’d estimate that the top 1/10 of 1% score all 99’s, meaning 45 kids nationwide. Even if the “99th” percentile extends to those scoring in the top 1.499%, there are only 675 kids nationwide (all grades, both sexes) scoring in the 99th, and perhaps 67.5 with all 99’s. Divide that number by at least 3 for any given grade, and in half again for one sex. I suspect people exaggerate their and their kids’ scores all the time.</p>
<p>Edited to add: I am confused about the hypothesis here. Are folks conjecturing that the total number of test-takers is not evenly distributed across all the percentiles this year? For example, are you hypothesizing that there are more 5’s than 9’s? If 3 wrong used to be a 99 (during the norm-reference period), but now everybody gets more than 3 wrong, then nobody (not literally, but you get the idea) gets a 99? Is that the idea?</p>
<p>Brother-
The 9th stanine covers the 96th to 99th percentiles, so if the 45,000 number is correct the number of kids scoring in the 9th stanine is more like 1800. Still a small number.</p>
<p>Brother, I am so glad you weighed in on this thread. I read with great interest your mathematical challenge to the assertion that it is common to find kids scoring in the 99th percentile in another thread. I thought, “At last! Someone is observing that the top 1% is by definition a really small number.” Thank you for that. Since you popped onto this thread, I would really like your insight on the ISEE scoring. I know from ERB that the percentile a child is assigned is based on comparing the number they got right versus the population who sat for the test the previous three years. The reason that I have heard scores were lower this year and last year is b/c last year the three years used to set percentile ranks were prior to re-norming and in the process of re-norming the test was made a lot harder. For instance, we heard that the number of questions scored reduced by a huge amount. In Reading for example, they only score 20 questions out of 25 tested and so the child can miss very few in order to get stanine 6 or better. Last year, if a 5th grader missed 1 question they were in stanine 7! Prior to re-norming there were a lot more questions in the Reading section and they graded more of them. When only 20 are being graded, you can’t miss very many. Also, I heard there were a lot of “Beta” questions in both last year’s and this year’s test and frankly it is just a harder test than it used to be. This year I would expect scores to be a little better, b/c 2010 is part of the pool. But since 2 of the 3 years are still pre-renorming years, again I understand it hurts the current test takers’ percentile ranks. Please weigh in. TIA.</p>
<p>EDIT: Another question I have concerns the bell curve. I know from ERB that the percentile ranks are not set based on the current year’s test takers. The child’s percentile rank is based on how they stand vis-a-vis the test taker’s from the prior three years. How does this affect the bell curve placement? Is the bell curve a three year bell curve? Would we expect most of the 8’s and 9’s to be from students who took the test in 2008 and 2009? Does this mean a child who received 7th stanine in 2009 be assigned stanine 8 in the 2011 bell curve, much the same way a student in 2011 would have had a 7 in 2009 but has a 6 this year? TIA.</p>
<p>Just want to give a real life example to all 9’s are rare. Dc took the ISEE test last year as 6th grader applying to 7th grade and received all 9s (99%, 99%, 99% and 99%). After receiving the score, I called ISEE and spoke with ISEE expert (who was a Statistician and oversaw the scoring and curve fitting process, etc). Without asking details, he told me that Dcs score is very rare and would put Dc rank No.1 or No.2 in the nation among 40,000 plus test taker. I am not sure what he meant by 40,000 test takers, is that all for 6th grader applying into 7th grade?</p>
<p>Sue, you are right in one sense. If the ISEE were to issue an overall percentile based on scores on all four subtests (relative to all other examinees’ scores), the 96th through 99th percentiles would correspond to the 9th stanine. But 9’s on all four subtests (even corresponding to all 96’s rather than all 99’s), which is what MiltonMom was talking about, would in all likelihood translate into the 99th percentile (i.e., the high end of the 9th stanine) overall.</p>
<p>Belkin39, thank you for your kind words! I am NO mathematician or statistician, so please take everything I say with a gigantic grain of salt. But I will try to address your questions: In theory, any subtest that has fewer then 100 questions (i.e., all of them) cannot have examinees falling into every percentile between 0 and 99. There has to be clustering (NOT a technical term; just my language). Here’s an example of a cluster: If fully 4% of testers get perfect (scaled) scores on Math Achievement, then a perfect score will be in the 96th (not the 99th) percentile. However, clustering may be mitigated to some extent by the different difficulty levels of questions given to different examinees. If there can be 100 different scaled scores (unlikely), there can theoretically be 100 different percentiles, though the scores won’t naturally distribute themselves that way anyway. (For example, the testers one year could (and this is a ridiculous example) fall into two groups: those that make no errors and those that get no answer correct; the actual percentiles would then be quite limited (to, say, 0, 50, 98 and 99, based on how hard each test form given to the perfect scorers was, with perhaps 50% falling into the 0th percentile, 48% into the 50th percentile, and 1% each into the 98th and 99th.) A more realistic example is the likelihood that (at least prior to changes in the test) examinees cluster around the middling percentiles (a la the bell curve you propose). Yes, someone scoring in the 70th percentile will have done better than 70% of students, but there may not actually be anyone achieving a percentile of precisely 70, and if there are some who score in the 70th, they may not represent a full 1% of examinees.</p>
<p>Compression of percentiles may be increased this year if the test was made harder but previous years’ norms are in use. If a scaled score above 900 used to correspond to the 99th percentile, but now almost no one can achieve 900+, then far fewer than 1% of examinees will achieve the 99th percentile. In other words, it is the 99th percentile in name only, and may not represent a full percent of test-takers. The same logic would apply to other percentiles. If the top 25% of scaled scores are clustering around the scaled score that USED TO correspond to a percentile of 50, those examinees will get 50th percentiles (or thereabouts) EVEN THOUGH they would be in the 75th+ percentile if the test were normed based only on this year’s scores. It will take a few years for everything to be correctly recalibrated. And I agree with you that fewer questions (that count) overall means more clustering, and I am pretty sure that this translates to lower percentiles for most, but am not positive. This is where we need someone better at math than I am. But your question is about the short run anyway: in the year or two or three right after the number of questions is lowered but old norms still hold, how are percentiles affected? I’m not sure, but may be able to think through the answer later. But I do agree that 2011 should be better than 2010 because of the rolling norms.</p>
<p>Phew. is there a mathematician on call who can help us?</p>
<p>I found the erb’s letter to families. The section below covers scoring: <a href=“http://erblearn.org/uploads/media_items/understanding-the-individual-student-report-for-families.original.pdf[/url][quote]”>http://erblearn.org/uploads/media_items/understanding-the-individual-student-report-for-families.original.pdf
</p>
<p>Brother and Periwinkle,</p>
<p>Thank you both for answering. Brother, my understanding is that yes, the norming group and the test they took will materially impact the percentiles if the majority of the norming group took an easier test. (And it was easier in those other years b/c there were more questions and less “beta” questions).-- I am not sure, but I think that it matters a lot that on ISEE there is no penalty for wrong answers. So the more questions you can get right, the better your score. The more graded questions available to answer the more chances you have to get more right. The main takeaway is that the dip in the past two years is significant enough that schools are very much aware of it. Periwinkle, what ERB fails to address in their letter attached to the score report is that norming is set against 2 out of 3 years where the children took an easier test, (more scored questions per section). They state that “A percentile rank of 63 on an ISEE test section, for example, indicates that the student scored as well as or higher than 63 percent (and lower than 36 percent) of all students in the norm group. Again, it is extremely important to note here that the group of students at this grade level who took a form of this test over the past three years is quite different from the group of students who took the standardized test the student may have taken as part of a large- scale testing program.” But what ERB does not state is that it is also a big factor that the test in 2008 and 2009 was very different from the test in 2010 and 2011. And it certainly has impacted the percentile rankings in a negative way. It would have been helpful for ERB to articulate the fact of the change in the test in the letter. Very few parents are aware of this and might feel a little freaked out by lower stanines than what they were expecting based on their child’s CTP performance or SATs. Luckily, the schools are aware and that is the most important thing.</p>
<p>In the stanines as listed above, the 4th to 6th stanines cover the 23rd to the 76th percentiles, the middle 50% of the distribution. If the test is correctly constructed, we should expect the average ISEE test taker to fall into the center of the bell curve. We should expect “lots of 5s.”</p>
<p>Anytime one introduces a new version of a test, scores should fall. This phenomenon is well known. If a subset of the test takers have been prepped to the prior version of the test, introducing a new version should erase that subset’s artificial advantage over the rest of the field. As the tutors learn the features of the new version of the test, this leveling effect will become less pronounced in coming years.</p>
<p>Measuring Up: What Educational Testing Really Tells Us, by Daniel Koretz is worth reading for a discussion of these issues.</p>