@Crusoemom – i think you mean 1440 - SI 218?
updated testmasters web site. check it out.
based on analysis of 10,000 students, raising Texas prediction to 219.
Care to post a link? I don’t see any updates…
10,000 is large, but was actually hoping for a larger sample set.
They estimated 60 of the 10,000 would be SF. Isn’t 60 a bit low? I show Texas having about 8% of total SF last year (1353 out of 16227). Someone said they called NM and was told Texas allocation would fall to 1300 this year. That would then be 1300 / 16227 = 8%. Wonder why they went with 60 which is 6%?
Going with 6% instead of 8% is significant. 8% would have lowered the cutoff below 219. Maybe back to 217. Thoughts on this?
Not based on a random sample. Again skewed to good test takers and those who studied.
Well, looking at their chart, 8% would probably push the cutoff to 218.
@itsgettingreal17, they show the peak around 118 SI. Thinking Applerouth showed the mean increasing over 214. So data does seem slightly skewed.
Hmmm… Interesting. If TX has 1353 SF allocation based on graduating seniors, and appx. 200K juniors take the test, their supposedly unbiased sample of 10K students should represent cutoff of the 67th student and not 60th. Would that land up being 218? Who knows?
Last year 227000 juniors took the test in Texas.
(1353 / 227000) * 10000 = 59.6
60 is right on the money.
Looking at the chart, 218 has around 20 students? So 60 for 219 or 80 for 218. From eyeballing it.
The problem is one of fat-tail distribution, they have about 15 students parked at 218 SI in their sample of 10K students. If sample is unbiased and for the state this lands up being 20 fold, you will have 300 students at that SI. How does NMSC plan to adjust for cutoff for the danglers?
Very FEW 225+ scores. As I said, number of Perfect and Near Perfect wouldn’t grow exponentially over past years. Maybe didn’t grow at all. No way, any states are going to be 225+ for cutoffs.
The right, high score, skew is very evident in the graph. Now, whether this is based on his sampling method or the actual data, I can’t tell. (Obviously, there is some bias because you expect testmasters’ clients to be right shifted. But how many datapoints of the 10K are testmaster clients?)
Based on the graph, it does look like the test was “too easy” to produce a normal (i.e. gaussian) distribution. Under these conditions it looks like the “most careful” test takers may have an edge over the “strongest” test takers.
@DoyleB, yep for 60 that is the right way to think about it. Thanks.
The distribution is strange. The mean according to CB expected to be at 148. The peak is well below that, with a lot of skew to the right. And there’s the bump around 218 that I hypothesized a week or so ago. The whole histogram is maybe due to a weird sample here, or maybe the nationwide distribution will look like this. Who knows?
@Plotinus - more evidence that good students do well, poor students do poorly, and a hole in between them?
I am counting appx 156 students at 210 SI or higher (1.56% of 10K students) - not compatible with top 3% of students needed at commended level and their suggested cutoff of 210 SI. With this sample, one would think commended level to be closer to ~205.
If this sample is heavily biased with students who paid to prepare for the PSAT, I’m not buying it.