My thoughts on that testmasters page:
This is horribly waaaay too simplistic and reflects an inherent lack of grasp of the distribution of the data.
Here are my thoughts on the College Board’s PRELIM concordance table. Refer to the table below.
As explained in my post here:
http://talk.collegeconfidential.com/sat-act-tests-test-preparation/1850258-possible-corresponding-range-of-nm-selection-indices-vs-2015-psat-scores.html#latest
there is not a one-to-one correspondence between the 2015 PSAT score and 2015 S.I.'s, because 2015 PSAT score calculation gives twice as much weighting to the M sub-score which results in different S.I. combinations.
I’ve merged the College Boards concordance estimate (the 2 black columns on the left) with my own calculation of possible range 2015 S.I.'s for a given 2015 PSAT score (the blue columns on the right).
**CB concord GMTplus7 calculation
2014 | 2015 || 2015 PSAT & S.I. | ratio of 2015 S.I. / 2014 PSAT score **
240 | 1520 || 1520 | 228 - 228 | 0.950
239 | 1520 ||
238 | 1520 ||
237 | 1510 || 1510 | 226 - 226 | 0.954
236 | 1510 ||
235 | 1510 ||
234 | 1500 || 1500 | 224 - 226 | 0.957 - 0.966
233 | 1500 ||
232 | 1500 ||
231 | 1490 || 1490 | 222 - 224 | 0.961 - 0.970
230 | 1490 ||
229 | 1490 ||
228 | 1490 ||
227 | 1480 || 1480 | 220 - 224 | 0.969 - 0.987
226 | 1480 ||
225 | 1480 ||
224 | 1470 || 1470 | 218 - 222 | 0.973 - 0.991
223 | 1470 ||
222 | 1470 ||
221 | 1460 || 1460 | 216 - 222 | 0.977 - 1.005
220 | 1460 ||
219 | 1460 ||
218 | 1450 || 1450 | 214 - 220 | 0.982 - 1.009
217 | 1450 ||
216 | 1450 ||
215 | 1440 || 1440 | 212 - 220 | 0.986 - 1.023
214 | 1440 ||
213 | 1440 ||
212 | 1430 || 1430 | 210 - 218 | 0.991 - 1.028
211 | 1430 ||
210 | 1420 || 1420 | 208 - 218 | 0.990 - 1.038
209 | 1420 ||
208 | 1410 || 1410 | 206 - 216 | 0.990 - 1.038
207 | 1410 ||
206 | 1400 || 1400 | 204 - 216 | 0.990 - 1.049
205 | 1400 ||
204 | 1390 || 1390 | 202 - 214 | 0.990 - 1.049
203 | 1390 ||
202 | 1380 || 1380 | 200 - 214 | 0.990 - 1.059
201 | 1370 || 1370 | 198 - 212 | 0.985 - 1.055
200 | 1370 ||
199 | 1360 || 1360 | 196 - 212 | 0.985 - 1.065
198 | 1360 ||
197 | 1350 || 1350 | 194 - 210 | 0.985 - 1.066
196 | 1340 || 1340 | 192 - 210 | 0.980 - 1.071
195 | 1340 ||
194 | 1330 || 1330 | 190 - 208 | 0.979 - 1.072
193 | 1320 || 1320 | 188 - 208 | 0.974 - 1.078
192 | 1310 || 1310 | 188 - 206 | 0.979 - 1.073
191 | 1300 || 1300 | 186 - 204 | 0.974 - 1.068
190 | 1300 ||
189 | 1290 || 1290 | 186 - 202 | 0.984 - 1.069
188 | 1280 || 1280 | 184 - 200 | 0.979 - 1.064
187 | 1280 ||
186 | 1270 || 1270 | 184 - 198 | 0.989 - 1.065
185 | 1270 ||
184 | 1260 || 1260 | 182 - 196 | 0.989 - 1.065
183 | 1260 ||
182 | 1250 || 1250 | 182 - 194 | 1.000 - 1.066
181 | 1250 ||
180 | 1240 || 1240 | 180 - 192 | 1.000 - 1.067
179 | 1240 ||
178 | 1230 || 1230 | 180 - 190 | 1.011 - 1.067
177 | 1230 ||
176 | 1220 || 1220 | 178 - 188 | 1.011 - 1.068
The assumption I made in post#2 about a simple 228/240 (= 0.950) scaling (using my 2 kids’ 2013 & 2015 results as benchmarks), only holds at the far ends of the scores distribution tail. As we work towards the middle of the scores distribution, the ratio of 2015 S.I. vs. 2014 S.I. does not have a constant 0.950 ratio-- if you plot the first two columns of black-font data in Excel, you can see that the concordance correlation is not linear. This is a reflection of the 2015 test being an inherently different test than previous years’ tests due to: no guessing penalty; no obscure vocab; a different range of math topics; lack of prep resources.
What this means in less geeky terms is: if you made a bellcurve barchart showing the frequency of student scores, the 2015 bellcurve has a different shape than the 2014 bellcurve.