Quick question: Lets look at the concordance of 2015 SI 210 since Testmasters has this # as commended cutoff per @firstrodeo’s table:
210 203 214 208 208 207.9 208
So this has a big spread 203-214, with mean around 2.8. So is fair to say, concording to the extremes (203 or 214) is possible but the LEAST probable outcome while the median of 208 is both possible and the MOST probable outcome?
I’m not a stat’s guy, so I’m not sure about probability or frequency of recurrence with sort of test, maybe that only applies to flipping a coin. But if it can be applied, then wouldn’t we have a decent predication now of the commended cutoff (SI 208) and start triangulating from there? That said I know a 208 SI commended cutoff makes no sense w/ the p11 2015 tables but then we know there’s something weird them anyway.
So some people are posting that they hear there are more 99%'s at their schools than in previous years and think that is some kind of proof of something. TOTAL NONSENSE.
When the state summary reports are released we will have an opportunity to see a distribution for every state and DC. That’ll clear things up, right? [-O<
At least it’ll be based on actual data.
It will depend on how they break this data up into intervals but if there are a bunch up at SI’s of 215-220, as it’s looking anecdotally and as @DoyleB is positing, then we’ll see some pretty interesting percentages up in the 700’s - 740’s. They can’t hide that - either the actual data will replicate in some way the SI table or it won’t. We should be able to see that and compare the anecdotes to the distributions.
@Mamelot wrote: “At least it’ll be based on actual data.” I sure hope so.
The intervals, with numbers, will tell us a lot. You are correct about that for sure. I hope they are broken up into bins that are small enough to be useful.
@DoyleB, to be honest, I don’t know why CB couldn’t smooth out the local peak you think might exist at 217-219 by adjusting the curves. But I may be missing something?
@exeover, if you have been reading the posts, you will know I’m in the camp that believes the SI% table will turn out to be correct. Time will show. CB had to know, you, me and many people would look at the 99+, 99, 98 regions of the tables to see where our students lined up. And they purposely removed the footnote that existed in the 2014 Guide. Keeping that footnote would have given them an opportunity to in some way lay out the parameters of the table and they choose to not do that (basically, they skipped the opportunity to have a disclaimer).
Regarding percentile inflation. Such inflation does exists, particularly in the National % numbers. As far as the User % numbers go, there was the change in definition from A to B. But @DoyleB and @thshadow have convinced me that is only 1 SI number, so somewhat minor.
True Believers. The CB SI tables on p. 11 are accurate and can be counted upon to predict 2017 NM cutoffs. Best arguments: CB has more information than we do, a long history with this test, disincentives to mislead, a large research study to work from, and the ability to set the curve wherever they like. They carried out statistical analysis on every individual question (to the point of tossing one out as unscorable). They knew they were under extra pressure to give accurate tables because this year was a new test. They labeled the concordance tables “preliminary” but not the SI tables, seemingly making a distinction between the reliability of the two.
True Pessimists. The CB SI tables on p. 11 are totally inaccurate and cannot be used for anything. Best arguments: 1) if you try to use them, then you wind up with highly improbable historical results along the lines reported in Cobb Co. Georgia and Illinois, e.g. here:
(http://talk.collegeconfidential.com/discussion/comment/19234638/#Comment_19234638). Perhaps some of the anecdotal evidence can be dismissed as “outliers” but cumulatively they are hard to dismiss. 2) The Compass education group & Applerouth analyses identify a lot of serious issues with the CB reporting of 2015 results. CB innovated rather wildly – with scoring ranges, definitions & categories of “percentiles” etc. – in order to align with SAT scores and seek market share from ACT, resulting in uncertainty or chaos in reporting the top 1%. 3) The SI tables are not based on the 10/15 test-taking population.
Fence-sitters: The CB SI tables on p 11 may not be 100% accurate, but they’re not arbitrary nonsense. They represent an intriguing mixture of potentially interpretable data.
@Speedy2019 Adjusting the curve allows you to shift or push/pull on the distribution. You can squeeze it, or spread it apart. You can linearly move it. So you can move the bump, or make the bump wider or narrower. But you can’t make it go away.
@MomNJof2, I had an excel spreadsheet with a lookup table that had an entry wrong. Fixing that. But, I now show NJ with 216-223.
For your case: “At the other end, a 225 could be generated with an 80 W, 70 R and 75 M. This would generate 38/37/37 for an SSI of 224.” I actually show this as 38/36/37 = 222.
I now show at the high end, a 223. (70 R, 77 W, 78 M would generate 36/38/37.5 = 223).
Do you agree 216-223?
This is a good check for my spreadsheet. If you agree, I’ll post an updated Concorded Range.
@DoyleB the goal will be for the tables NOT to be useful the way we’d like them to be. And they can certainly hide a little bump (‘little’ from the perspective of the overall curve) in a wide enough interval. And they can hide a bump in smaller intervals simply by putting half in a lower and half in an upper. There are many things they can do. But I think we’ll still see something revealing because the state distributions won’t look like they have historically if there’s some weird clusters of high scores this time around.
We would like to believe the SI percentiles reported by CB in Understanding Your Scores are accurate but,
there are a number of reasons for doubt - only time will tell for sure.
In the Understanding Scores doc (https://collegereadiness.collegeboard.org/pdf/2015-psat-nmsqt-understanding-scores.pdf) it does say: “The percentiles accompanying students’ scores have been derived from a research study.” So not actual Oct. 2015 test takers. Given that I do not think the CB is actually reporting the “true” percentiles that each SI equates to based on the 1.7 mill of real juniors who took the PSAT. I am not sure why they are being less than clear but many specualte it is to encourage to take the SA.
"The CB concordance charts say “preliminary” and are to be revised re-released in May. About Concordance Tables and their purposes, the CB "Understanding Doc says: "Key uses for concordance tables There are several reasons why students and educators need to compare scores on the PSAT/NMSQT from 2014 and earlier to the redesigned PSAT/NMSQT (2015 and future.)
. . . .
» To understand approximate NMSC Selection Index scores on the redesigned PSAT/NMSQT for
students who took the PSAT/NMSQT in 2014 or earlier. (Concord NMSC Selection Index scores from the PSAT/NMSQT from 2014 and earlier to the redesigned PSAT/NMSQT [2015 and future].)
And later it says "NMSC Selection Index — For guidance purposes, counselors can approximate an
NMSC Selection Index on the redesigned PSAT/NMSQT from PSAT/NMSQT scores from 2014 or earlier. Use the
Section to Test Concordance Tables on pages 26–27 to convert the three subject test scores. Then, double the
sum of the Reading Test, Writing and Language Test, and Math Test scores to calculate the approximate NMSC
Selection Index " I am still not sure what this means since an actual SI is on each student’s report - problem is it does not always match what is derived using the 3 scores and the concordance tables.
So, look at GC reports and as much real test taker data has been helpful as well as thinking about ranges of SI’s that might fall within certain percentiles. No one knows for sure of course what the percentiles really are for each SI or the state cut offs, except CB or NM. State summary reports – hopefully out in Feb – and revised concordance & leaning the commended cut off score in April or May should help us get a better sense.
@thshadow Thank you for your explanation in post #2175 about how one could arrive at 214. Why didn’t I think of that!
@AJ2017 came up with an idea that instead of actually calculating percentiles, maybe CB just guessed at what they ought to be based on a normal curve and using the mean and standard deviation. I thought that I had disproved this by calculating that, if they had done this, then the lowest 99+ (which I am assuming means 99.5, but I could be wrong) percentile would equal 215 and not CB’s figure of 214. But the shadow explained that since the mean and standard deviation figures are rounded off, one could possibly get 214, depending on what the non-rounded figures were. I think, however, that I can disprove the normal curve extrapolation theory by checking out the lowest score for the 97th percentile, which I think was CB gave as 200. I believe if CB were just extrapolating from a normal curve they would have given it as 197 (or in the range of 195 to 199 depending on different guesses about rounding.) The z score for 97th percentile is 1.8808
Thanks @Speedy2019 – appreciate it. Think this is they helpful and of some more data comes in then perhaps the
last 2 columns can be tweaked: Liberal Cutoff…Conservative .
But maybe at the 97% they had more real data from the research study. It is only at the top 1% where they had a smaller group of scorers and maybe more funny bumps that they were unable to correct for in how they scored the test so they reported on page 11 what would be expected with a normal distribution. They would be better able to describe what happens in this top percentage when they have a larger population. Like next year when they can use last years testers as a reference. It is just a theory. (Believe me I have not thought about statistics for a very long time if ever really). I know it would seem irresponsible for CB to do that. If what I am proposing is the case why would they feel they have to specify 99+% on page 11. Why would not they just leave everything 205 and above 99%? I don’t know.
I’ve been checking this board frequently because I’ve been worried about my DS’s 218 making the cut in Missouri. But it looks like I may have an additional worry- the commended cutoff.
From what I’ve read about previously qualifying for NMF, you merely had to take the SAT as a formality. The score you had to make was basically equal to what commended was, and that was fairly far below the cutoff in most states.
But if commended is indeed quite a bit higher this year, it may be too close for comfort for a LOT of students. (Of course if a lot of students felt they needed to take the SAT more than once that would only help College Board’s pocket books.)
We need to be studying for the ACT now, and our time may be divided between tests.
So if commended were 210 or 211, what do you think he would need on the SAT in March to qualify for NMF?
Congratualtions @rb681000. Ac score of 218 in Missouri is a great score and very very safe from what have been discussed so far here. If somehow the commended is 210, the corresponding SAT score is simply 2100 in older SAT format, that would translate to 1400 out of 1600? Anyway someone who scores 218 would be expected to find SAT easy enough to get 2100 (old sat), I say this based on my SI score and SAT score achieved in 2015 SAT exam.