Dear colleagues:

My thanks to everyone who participated in the meeting today.

As I left the meeting the Language and Literature chair passed me page 86 from the book "Six Practice TOEFL Tests" and indicated that this conversion table was more accurate than the one provided to me on the weekend.   I gather the chair did actually fall down on a trail on Saturday injuring his foot and hitting his head hard enough on a tree limb or some other blunt object such as to cause a concussion and a trip to the hospital.  The chair appears to be all right except for a limp.  I will presume that the second wrong table I was given is related to his condition on Saturday.

I have now spent another evening crunching numbers with my family wonders when it will all come to a final conclusion and they might see something other than the back of my head in front of a computer.  My wife just looks at the screen and says, "Wasn't that what you were doing the other evening?"

I have learned something valuable: the TOEFL test is, for our students, highly sensitive the exact number one picks as the cut-off.  Moving the number by one or two correct on the raw score shifts dozens of students.

I have also learned that the gap I need to worry about is not the vertical gap between my least squares (best fit) linear regression (line) but the horizontal gap.  And the horizontal gap at TOEFL 400 is large: 2.59 points.  This creates an actual difference of 3 points, which makes for substantive changes in the numbers.  As a result I have abandoned the linear regression and gone to using the tables directly.  I have used a special function in spreadsheets called VLOOKUP to look up the correct TOEFL conversion from the table.  This means that now some scores that sit in the middle of a range go to the bottom of the range.  Both a 30 and a 31 on the reading section are now a 460, while a 32 is a 470.  So there now is no difference between a 30 and a 31.

The hour is late and I have an 8:00 A.M. final tomorrow, so here are some of the other statistics I have generated this evening:

The new cut-offs (minimum raw scores):

 Natl struct 24 Natl read 32 Sum 56 IEP struct 16 IEP read 21

Number of students impacted:

 Changed sum 206

The new distribution of students:

 Number code Count Location 1 465 State 2 383 IEP 3 511 National

The following is the old distribution in the leftmost column and the new distribution in the bottom row:

 Count - Newplace Newplace ProjDest 1 2 3 Total Result 1 440 440 2 23 202 225 3 2 181 511 694 Total Result 465 383 511 1359

440 were assigned to certificate programs and are still assigned to certificate programs.
225 were assigned to the IEP, that number grows to 383.  181 leave group 3 ("national") and drop down to IEP.  IEP, meanwhile, drops 23 students to the certificate level.
694 were assigned to group 3 ("national" or post-IEP), that drops to 511.  Note 2 students are dropped down to the certificate level.

Note that I reported in the meeting that 355 students were affected, the new new table affects only 206.  Next year I would ask that the official table be turned over to Ringlen at the time the test is submitted because the results are so sensitive to the table used.  We MUST abandon the use of the TOEFL.

The faulty table (the first one) generated the following distribution by high school:

 Count - ProjDest ProjDest HS 1 2 3 Total Result Berea 10 5 9 24 CCA 9 9 CHS 178 50 40 268 CSDA 3 1 9 13 KHS 10 27 89 126 KSC 2 1 6 9 Mizpah 3 8 11 22 NCHS 9 1 1 11 NICHS 7 8 14 29 Ohwa 2 3 5 10 OIHS 3 4 21 28 PATS 1 3 20 24 PICS 19 40 266 325 PLHA 19 12 7 38 PPSD 1 22 23 PSDA 1 21 22 SCA 3 6 33 42 SNHS 66 16 7 89 Weno 87 18 8 113 Xavier 26 26 YHS 16 20 67 103 YSDA 2 3 5 Total Result 440 225 694 1359

The new new table generates the following new distribution:

 Count - Newplace Newplace HS 1 2 3 Total Result Berea 12 7 5 24 CCA 9 9 CHS 185 67 16 268 CSDA 4 3 6 13 KHS 13 42 71 126 KSC 2 4 3 9 Mizpah 3 13 6 22 NCHS 10 1 11 NICHS 7 15 7 29 Ohwa 2 5 3 10 OIHS 3 11 14 28 PATS 1 5 18 24 PICS 23 98 204 325 PLHA 20 15 3 38 PPSD 7 16 23 PSDA 5 17 22 SCA 3 16 23 42 SNHS 67 18 4 89 Weno 90 21 2 113 Xavier 26 26 YHS 18 29 56 103 YSDA 2 2 1 5 Total Result 465 383 511 1359

The directors should be able to work from the above table to determine their new IEP numbers.  I reported to Director Kephas in the meeting that 64 (not 89) would be eligible for the "national" campus (post-IEP), but the new new table puts 71 into the "national" campus.  I use quotes because Director Weilbacher rightfully reminded me that degree programs such as HRM are offered at Pohnpei State Campus.  Off-island students can dorm in Palikir and attend HRM at Pohnpei campus.

The new new table also ups the number of IEP students from 32 (27 under the original faulty table) to 41, principally at the expense of the certificate programs.  Only 13 are eligible for certificate programs (10 under the original faulty table).

As I noted earlier, no matter what table is used, underlying improvement in teaching at KHS seems to have an impact there.

In Yap, YHS + OIHS + NICHS + YSDA  yields an IEP of 57 or so.  Likewise the other states can tally the new numbers and then make their own adjustment for the known acceptance rate.

The TOEFL conversion I am now is below, with column one the raw score and column four the most recently received table (one might note that the May 10 and May 12 tables are very different in certain ranges):

 Raw Converted TOEFL 0510 TOEFL 0512 0 20 200 210 2 22 220 220 3 23 230 230 5 24 240 250 6 24 240 260 8 25 250 270 9 26 260 280 11 27 270 300 12 28 280 310 14 31 310 330 15 32 320 350 17 35 350 360 18 36 360 370 20 38 380 390 21 39 390 400 23 41 410 410 24 42 420 420 26 43 430 430 27 44 440 440 29 45 450 450 30 45 450 460 32 46 460 470 33 47 470 480 35 48 480 490 36 49 490 500 38 50 500 510 39 50 500 510 41 51 510 530 42 52 520 530 44 53 530 540 45 54 540 550 47 55 550 560 48 55 550 570 50 57 570 590 51 58 580 600 53 59 590 610 54 60 600 620 56 64 640 640 57 66 660 660 59 67 670 670 60 68 680 680

If any member of the admissions board would like the whole spreadsheet, let me know.  But be forewarned: it is 4 megabyte spreadsheet.  That is large.  Older computers might simply refuse to run it.

- Dana