I saw that too @Mathyone. I’m going through the file structures here that are used to feed the data. I don’t see how @gasenioryear 's “PERCENTILE OF EACH STUDENTS SI NATIONALLY” is any different that the SI user percentile’s we are seeing on the CB site.
In the file I see these columns referencing SI percentile (at a total level, then at the section level - which doesn’t really matter):
PERCENTILENATREPPSATTOTAL (this the National % we see on the report for SI)
PERCENTILENATUSERPSATTOTAL (this is the "User SI % total, and they also show each section %, which we can see too on the CB site).
SELECTIONINDEXPERCENTILE (perhaps this is the field?) If we know the component level user %, why is this field more helpful?
The percentiles I see on the reports are for the subscores in each section and then the combined reading/writing and Math and the total overall score but not for the Selection Index number. As we know the same total score can result in a range of SI numbers and if there is some official indication of what SI’s are in the 99th%ile. as I think @gasenioryear is indicating here – http://talk.collegeconfidential.com/discussion/comment/19165844/#Comment_19165844
as being an SI of 207 that would be a useful benchmark to guesstimate how scores above 207 might range above the 99%ile - needed for the states with the higher cutoffs. If any more info can be gleaned from reports the GCs get, please share it as you learn of it. Thanks!!
@CA1543 I get it now. If the lowest 99% last year was 213, and it appears (from @gasenioryear 's sleuthing) that the lowest 99% was 207 this year, then it appears that cutoffs WILL INDEED be lower than last year, but about 6 points. This makes sense and is awesome!
It is just that the user %iles CB is providing is “% of a typical group of 11th grade” not the ACTUAL juniors who took the test. Since for cutoffs we need 99.3%-99.5% data, depending how large CB’s typical group was, the estimates can be quite off for those small higher scoring range. So it is rather uncertain at this point.
@payn4ward – That is a good question - are the GC’s getting any “real” data based upon actual test-takers. If not then the “207” SI is just a guess - but the College Board has the actual data at this point so not sure why it can give more accurate percentile info nationally, if not state by state. I would think States will need to get this info to make their allocations – I understand each state gets a certain allocation of NM semi-finalists based upon their number of graduating seniors as a percentage of the total graduating seniors in the US (so not tied to number of PSAT takers apparently). If there is a hard cut-off on a number of semi-finalists at or around 16,000, then if a state has a large group at a particular SI & thus can’t include them all bc it would go over the allotment, maybe it is forced to award less that its allotment (a state can’t exclude some students with the same SI, can it?)- not sure how that works or its implications, Maybe the 16,000 is an estimate or there is a bit of leeway to States?
I don’t believe there are any state organizations involved. The National Merit corporation identifies the NMSF using data provided by the College Board and graduating class size data obtained from government and other sources. So, the National Merit organization works out the details of making the final number close to 16,000 by adjusting state cutoffs. All students at the cutoff number in a state are NMSF.
Thanks - understand states having nothing to do with their cut off scores - must just provide data about number of graduating seniors they have to NM folks?
@suzyQ7 I’m just quoting CB’s explanation next to their “user percentile”
“Your percentile indicates the percentage of a typical group of 11th grade College Board U.S. test takers who would have had scores at or below your score.”
I got a 1500 (SI = 224). I recently moved to IL so I have no idea how competitive my score is for this state. Would I be guaranteed NMSF or am I borderline?
They do “focus group” testing. But the numbers of test takers are small. (And, some kids have reported that “no one takes it seriously” because it doesn’t count.)
Its ridiculous if they didn’t use actual numbers when they HAVE them. Its hard for me to believe that is true- they are probably just saying that to cover themselves.
If you click the little “i” icon next to the percentiles, it says the “percentiles are derived via a research study sample of U.S. students in the 11th grade, weighted to represent …” So, all the percentiles that we are reading so much into are based on a sample that is clearly much smaller than 2 million kids (and less motivated to do well).
It is likely that the preliminary concordance tables are also based on the percentiles from this “research study” (though some people see funky differences indicating they may be based on different data sets). The actual percentiles that come out (April?) could be very different from what the score report shows. In fact, I would not be surprised if 120 and 121 (maybe even 122) turn out to be not a sure thing in all states. I still agree with earlier speculation that they made the test too easy at the high end with not enough granularity.
I think it’s important to remember that the SI is out of 228 this year, not 240. Projected scores from the concordance table seem to relate more to being out of 240 than out of 228 because they seem so high. (although I think people have already established that the concordance table predictions seem a little weird).
I have been thinking of it as a proportion in a way. Would it be wrong to look at it like this?
For example, Louisiana’s 2015 NMSF cutoff score was a 208. The percentage of a 208/240 is about 87% rounded. Now, if you change the SI to be out of 228, a similar score would be about a 198/228. That is about 87% rounded.
This puts a proportionally similar score as way lower than what the predicted cutoff from the concordance tables. This also is similar to the sliding cutoffs predicted by test master. Any thoughts?
Other things to consider are the easier test (they’ve described it as having a lower ceiling), and the lack of penalty points, which will both tend to “raise” cuts (from the -12, not necessarily from last year’s cuts).
Where has it been established that this year’s test was easier than last? We are getting mixed reviews. And it doesn’t appear that juniors at my daughter’s school did better than last year’s class.
I am guessing that the test will have a “flatter” bell curve, with both more students scoring on the left and right tails. The non-calculator math was probably very challenging for some students who have been raised with calculator-dependent teaching. The lower ceiling in item difficulty will mean higher scores for other students.
Does anyone know whether more students in the US took the 2015 PSAT exam as I heard the numbers were much higher which would impact both the number of NMSF recipients and to the extent that the overall curve is flatter would decrease score required to achieve same?
Looking at a 219 in NJ so trying to figure out what it could mean.
Hasn’t College Board outright advertised this as not having enough head room to hit 800 in any section? And that when in years past sooooo many 80’s, especially in math. It boggles my mind a bit. Those top math students, the ones who nailed algebra in elementary school, and going to blow that curve completely.