Last 4 years for HISD:
2016 - 63 SF
2015 - 56 SF
2014 - 60 SF
2013 - 56 SF
Pretty consistent.
@PAMom21 - Reading comprehension seems to be a huge problem in general (outside PSAT/SAT/ACT). I am always surprised at schools with only 33 or 49 percent of students who read at grade level. Was this always the case? When did such low comprehension happen?
Can only imagine what it means for wordy math problems and how to teach to extract the number problems from the text.
This is so spot on that its worth repeating:
âWow, this TX data looks extremely different from the sample NMSI percentiles the College Board posted. ** It appears that the top 3% of test takers did better than expected, thus the NMSF cutoff scores are higher than anticipated.** However, on down the list things are not looking good for the class of 2017. The CB expected the median score to be 146, but instead of 50% of students scoring a 146 or better, only 31% of the TX sample did. This sample didnât reach the median until a score of 130. ** Basically the scores 200 or above meet or exceed expectations, but the scores below that are poorer than expected.**â
This is exactly what @Plotinus predicted had happened and fits into the anecdotal evidence on this site. The really smart kids thought it was easy. On the other end of the spectrum, this test was harder.
I am rooting for you too @srk2017 !!! Test Masters put CA at 219 so even if goes up a point your D is ok. WE wonât know of course till September. Wonder if any more state or city data will trickle out though as time goes on.
@Speedy2019 and @DoyleB - I think itâs not just at 218 (actually I might argue itâs 216), there are many non-linearities. For example, thereâs another hump at about 224 or so, and one at 202, etc⊠Also, why from 130 to 138 do you get more and more kids, despite the fact that this is after the peak? Itâs almost like a sawtooth wave is mixed in the whole distribution, which is rather weird⊠???
I donât have a good sense of how weird a real set of (normally distributed) test takers should actually look though.
Sawtooth is to be expected because SIs result from doubling (and then adding) scaled scores. Only some math scores (those with a .5) even allow for an odd SI.
@MatzoBall â thank you - appreciate your comment on sawtooth and why â I knew this somewhere in my mind but what you are saying just gave me a V8 moment â thatâs why perhaps 218 gets cut off and 219 - more rare â makes it in Texas, NY etc. Perhaps the effects are similar elsewhere but at different odd numbers towards the top? Odd numbers are fewer arenât they in terms of students who land there - easier cut to make? What do our stats experts think?? Heck, I suppose CB might actually structure the test for this effect.
Now because of this difference between a Odd score and Even score, it will be now more difficult to allocate exactly 1300 #NMSF to Texas or 2000 to CA, whatever be the allocation. Moving one score up and down will make the number of students change quite a bit. CA 220 now become even more source of anxiety.
I can see that if the pattern was odd vs even, or even every 3⊠But it seems to be over much larger ranges than thatâŠ
So I agree that itâs part of the explanation - but is it all of it?
A few thoughts:
(1) Testmasters refers to its prediction as Texas being âno higher than 219.â Perhaps theyâre thinking of a range, say 217-219 (although in an e-mail exchange, they say that the actual TX score could be a point or two higher, but I think they would have come out and said 220 or 221 if there was any reasonable chance to so conclude, in order to hedge against disappointed Testmasters students).
(2) In another e-mail exchange after the TX was revised, Testmasters, interestingly, stuck to its prediction that Virginia will be 218 (or thereabouts). I think they said that, not to stonewall, but simply to refuse to comment on other states without having the data. Sounds like Testmasters is done making predictions, about TX or anywhere else.
(3) Based on the PSAT SI index, Testmasters analysis, anecdotes, and the terrific comprehensive analyses of Speedy, Doyle, Dallas, Shadow, and others, we may just need to content ourselves with ranges rather than a target number. For example (and for illustrative purposes only):
New Jersey 220-224
District of Columbia 220-224
Massachusetts 219-223
California 218-222
Virginia 217-221
Texas 217-221
Georgia 215-219 and so on.
Many of the anecdotes and data (Cobb, Testmasters) seem to favor the higher ranges, but Iâm not ready to dispense entirely with the SI index. As Testmasters notes, the data is pretty weird.
To me, the newly released chart looks like a combination of two much smoother distributions (each more normal): a big one centered around 100 and ended around 214, the other one much smaller but centered around 218-219. The reason I think so is because if you look at any previous yearâs distribution, the band for 97.0%-98.0% is narrower than that for 98.0%-99.0%, which is narrower than 99.0%-99.5%, which is narrower than 99.5% to the end. This is natural for distribution of any large enough samples. There are more combinations (in terms of getting which questions wrong) to get 97% than 98%, and so on.
The TX one is strange in that as DoyleB showed in #3027, there are 8 scores between 98.0% and 99.0% and 4 scores between 99.0% and 99.5%. For the previous two years, there were 7 scores for 98.0-99.0% each year, and 11 and 10 scores for 99.0%-99.5%, respectively.
With more data, I believe the curve will smooth out. If 217 as lowest 99% stays, the cutoff will be high. However if 209 as lowest 98% stays, the lowest 99% has to move lower, and so will the cutoffs.
I think this curve is based on two distinctively different populations. How the general population looks like is anyoneâs guess.
@AnnMarie74 with post #3049
â@PicoLA Last year, 223 was 99.59th percentileâ
Is there anyway, could you provide more info about 99.59 (link, docs âŠ)? Itâs is important since in 2014 PSAT, SI 223 is the highest slot in 99% range and there are 11 band slots in 99% range. Just for crunching number purpose, thank you in advance
@thshadow Maybe a bit of the sawtooth is related to the âlinkedâ questions in the reading sections. A lot of kids went either 0-2 or 2-2 & that might have contributed something to the bulges at the top.
Every time Iâm away for an expended time, a ton of news when Iâm back on with CC. Thanks everyone for all your thoughts, analyses, tables,lists, on and on.
@Speedy2019 Please share with us the data from the other test prep company once you have it.
The data for TX does behave strangely, a large disparity between the top 3% and the majority of the state. Perhaps this attributes to people who really prepped for the new test.
I just browsed www.testmasters.com for a bit. They do pride themselves in improving scores more than other test prep companies. They also have perfect scorers AND do prep for the PSAT. I would expect that their samples are indeed somewhat skewed towards the very top.
From their website:
âSince 1991, Testmasters has earned a reputation for offering quality test preparation products with our decades of experience. The proof is in the Testmasters Score Improvement Guarantee. Take the Testmasters course, score higher than you thought possible â itâs that simple. Depending on the test, our score increase guarantees are up to 50% more than our competitors. Over 160,000 students have taken our courses. Thousands of highly satisfied students cannot be wrong. Read more details about the Testmasters Score Improvement Guarantee for each test.â
http://cdn.testmasters.com/wp-content/uploads/2014/05/perfectscores.pdf
@Nathanb I think this curve is based on two distinctively different populations. How the general population looks like is anyoneâs guess.
I agree. It looks like maybe Testmasters combined a âhigh-scoringâ with a âlow-scoringâ group. This enabled them to 'get the right answer '(i.e. find the top 60 out of their sample of 10,000) for TX because they presumably have historical data on both TX groups and are confident in their combination. But if that was their method, we still donât know how real the âbulgesâ are. They might be partly the product of combining two different groups.
Does 215 for Florida seem likely for the cutoff?
@LadyMeowMeow Thatâs why Testmasters appears to be hedging their prediction by saying âno higher than 219â rather than simply pegging 219. Really believe this could be 217, 218, 219, or 220, with almost an equal chance at any of these four numbers.
I believe TestMasters mislead viewers with their data
Reference to post #3074 by @TallyMon2017
For example, I find Jessica Wu, who had SAT perfect score in 2002 as in the following link
http://www.yourhoustonnews.com/archives/test-masters-perfect-score-factory/article_18e8a6ce-c83e-54b0-b61b-a03dd8d321d7.html - use the Control F, then type Jessica Wu at top right box
From TestMasters Perfect Scorers list- http://cdn.testmasters.com/wp-content/uploads/2014/05/perfectscores.pdf
We have average 550+ perfect scorers every year. That mean for the last 25 years, we have at least 13,500 perfect scorers
Can we trust TestMasters, who would use 2002 Jessica Wu record, 13 years ago to advertise? I believe Jessica Wu is a real person.
They trained 160000+ students and produce only 90 perfect scorers in 25 years of doing business. I m not impressed
TAMS (North TX) has enrollment 360+ for 11th, 12 th grades, usually got at least 8 SAT Perfect Scorers every year
@SoccerLad I have no idea anymore 
But I really, really hope so!