The Nationally Representative Sample percentile shows how your score compares to the scores of all U.S. students in a particular grade, including those who don’t typically take the test.
The User Percentile — Nation shows how your score compares to the scores of only some U.S. students in a particular grade, a group limited to students who typically take the test.
So it seems as neither percentile is an actual percentile based upon people who took the test. The National Percentile is how they think you stand up against all juniors. The User Percentile is how they think you might stand up to the typical junior who takes the test. However, it does not appear to be an actual percentage to be measured based upon this years results.
Thanks @candjsdad. This chart is similar to what @BunnyBlue has predicted on various forum pages. The states with lower 2014 cutoffs will likely move to a higher cutoff number and the states with a high cutoff number will likely move to a slightly lower cut off number. The result is a flatter spread.
@candjsdad – thanks for posting the link to revised predictions of cut off scores by Test Masters - I had emailed them and shared the data collected on CC. Who knows what info they are using but at least there is apparently some further thinking behind their approach. We’ll learn more in due time I am sure.
" The states with lower 2014 cutoffs will likely move to a higher cutoff number " Lower scoring students will have gotten more questions wrong. In the old format, they would have deductions which really begin to add up. In the new format there is no deduction for wrong answers. (I am assuming that explains the strange low end of the math concordance). I would have expected the scaling that they do on the scores to take this into account, but perhaps not, and that could explain why the lower cutoff states might rise a bit. I don’t recall off the top of my head what the low cutoffs were in terms of questions missed, but if it was something like 20 questions, that’s 5 points not being deducted any more.
I am not sure why it is quite as flat and compressed as it is and some states are predicted to go up so much - the Concordance tables were preliminary and the change in the scale is not that dramatic. Things might change based on actual test results and info on test takers on a state by state basis. Also, some states have NM semi-finalists at a bit less than the 99%ile as I understand it (is that right CC’rs?) so an SI of 210 as the lowest seems high to me based on all the data collected on CC: http://talk.collegeconfidential.com/discussion/comment/19162923/#Comment_19162923
Once the cut off for commended students comes out - likely we’ll learn it in April - i think closer predictions can likely be made. But the revised predicted cut off scores by Test Masters seem at least more based on a rationale. Lets see what Prep Scholar does if anything in terms of revising their estimates.
For those that are surprised about the uptick in the number of students taking the test, remember that the SAT is changing; likely many wanted to test the waters to see if they wanted to go with the old format or the new format. Taking the new-format PSAT was the best way to accomplish that.
Also - it was only given on a Wednesday - not also on a Saturday in 2015 so I imagine that had some impact on the numbers of students who took it, though I can’t find a report of how many students - numbers of actual sophomores & juniors - who took the test. And as has been noted by CCer’s, the percentiles listed on score reports (for Total Score, Evidenced-Based Reading & Writing & Math) are not “actual” based on the real pool of test-takers in 2015, so it is hard to assign much real value to them or have much confidence at this point - at least not precisely. I would have thought having all the real raw data the College Board could have been more exact. But I do think the components in the report and expanded feedback will help many students as they consider SAT prep or whether to try the ACT.
Well, if the percentile data is just “estimated” (as it appears to be), then it’s logical to assume that the concordance table data is estimated as well…
Given that they’re both estimated presumably at the same time from the same data, though - why don’t they agree with each other?!? :((
I would read anything like this with +/-3 uncertainty.
The whole underlying assumption could be wrong, though.
That is, what CB provided has any meaningful info.
