Subject: Re: Did the SAT scale change a while back? Date: Wed, 20 May 1998 02:33:17 GMT From: nospam@spamaway.com (SteveK) Organization: EarthLink Network, Inc. Newsgroups: misc.education References: 1 , 2 , 3 , 4 I received an email from someone that explained to me that the SAT is a bell curve and that every so often it gets adjusted to keep it a bell curve. I suggested that the person who emailed me the explanation post it to misc.education where the original question started. He said that he couldn't post it easily so I volunteered to do so for him with his permission. He gave me his permission. The following is his explanation. (My question is has a ">" in front of it.) _____________________________________________ >My memory of my scores from many years ago was that there was one math >score and one verbal score. How does that fit in with a range of -10 >to 50 or 0 to 80? Are you saying that the math score changed by 60 >and the verbal score by 80 and that one of them went up and the other >down? > >krausteratearthlinkdotnet No, he is saying that the change in the scores was not a simple adding or subtracting to them. If that had been all that were needed there would have been much less justification. The concept in the scoring of these tests is that they reflect a bell curve statistical dsitribution with a mean of "100" (for IQ scores) or "1000" for SAT scores, and a standard deviation of 15/150. Thus the minimum SAT score of 400 respresents a score 4 standard deviations below the mean, which is so extreme that any attempt to measure more exactly is considered to be statistically unsound. (Thus those who say that merely filling in your name gets you 400 points are in error - it is merely that any score that would be below 400 points is reported as 400 points), and 1600 is NOT a perfect score, but merely a score so high that they will not raise the score higher merely for getting a couple more answers correct. Over the last few decades the average score in the US dropped into the low 900s. No conclusions are certain from this, since the test changes every year - for the most part, it is believed that the test has gotten slightly harder with time, but this is a subjective judgement. What is certain is that school curricula have changed and the SAT emphasis on abstruse vocabulary was not testing things that the colleges cared about (and the purpose of SATs after all is to tell the colleges which kids they should admit); they were more interested in the types of things that cause someone to need remedial classes, for example. Thus the test needed changing in some significant ways. In addition, in the 1950s when the scores were last renormed to 1000, only around 10% of kids took the SATs, mostly the top 10% of the kids in the schools. Now more than 50% of all kids take the SATs and in some states, closer to 90% of kids take the SATs (in the Plains and Mountain states, many kids take the ACT instead). Needless to say, the distribution of scores for the top 10% of the kids will be different and far higher than the distribution for the top 50% with large portions of the bottom 50% thrown in from some states. Different not only in that the average is lower, but also in that the curve is no longer a simple bell curve based on the old scoring system. As a result, the old scores were no longer statistically useful. You could not tell how much better a 1100 score was than a 1050 score, and it was certainly NOT the same as the difference between a 1050 score and a 1000 score. In fact you simply couldn't tell what the scores meant other than a broad "below average", "average", "above average" "way above average". The change in the curricula meant that verbal scores changed differently from math scores, so that summing the two scores (as is normally done) was even more fallacious. To cite an extreme, a score of 700 verbal 300 math (total 1000) no longer could be compared to a score of 500 verbal 500 math (also 1000). For the recentering then, they 1) redesigned the tests to deemphasize the vocabulary sections, increase reading comprehension difficulty and make the math test a little harder, and maybe some other things that I don't recall; 2) established a new Bell curve by giving this test to a population of students, and normalizing it so that the statistics worked properly again; 3) nationally switching to the new test style, and using the new bell curve scores instead of the old ones. They also calculated scores using the old norms so that people could see how scores changed under the new Bell Curve. The change to math scores ranged from a descrease of 10 points to an increase of 50 points, depending on where on the curve you were. The change to verbal scores ranged from 0 to 80 depending on where on the Bell curve you were. Obviously, the lowest scores in the country remained at 200 in either test area, and the highest scores remained at 800, so therefore the change to these scores was 0. The change around the median score that formerly averaged 920 probably was closer to the maximum change (but wasn't the peak change because a 50 point increase in math and an 80 point change in verbal would have made a 920 score into 1050 and not 1000 as it should be). To see the effect on a particular score, you need to look up in a table which Educational Tetsing Service did distribute. But given the stimultaneous change in test and scoring, it is not easy to figure out what any change "means", if anything. And ETS is not bothered by this. They not only don't care that SAT scores are no longer easily comparable to older scores, they have explcitly said that the DON'T WANT the scores to be used for such comparisons, since that is not their purpose, and conceding to that as a purpose would invalidate the true purpose of measuring kids for admission to college THIS YEAR based on the education that they have currently received. But this desire doesn't stop people from making such comparisons, since there is no other comparable national test, and people want to compare schools at the K/12 level to argue for their pet reform and/or decry the state of American education. It is just that most arguments based on SAT scores are worthless, especially if you don'tlook closely at what the scores represent. Hope this more detailed explanation helps you. I cannot post to newsgroups quickly so I am doing this by email. lojbab ---- lojbab lojbab@access.digex.net Bob LeChevalier, President, The Logical Language Group, Inc. 2904 Beau Lane, Fairfax VA 22031-1303 USA 703-385-0273 Artificial language Loglan/Lojban: ftp.access.digex.net /pub/access/lojbab or see Lojban WWW Server: href="http://xiron.pc.helsinki.fi/lojban/" Order _The Complete Lojban Language_ - see our Web pages or ask me. (end of posting) -------------------------------------------- I copied and pasted his email from Eudora to Agent and may have added some formatting mess with line feeds. krausteratearthlinkdotnet