Presentation at the Fall 2011 Meeting of the Michigan Educational Research Association Identifying MME Cut Scores 2 University 2-Year Institution 2-Year Institution Central Michigan University Alpena Community College Mid Michigan Community College Eastern Michigan University Delta College Monroe County Community College Ferris State University Glen Oaks Community College Montcalm Community College Grand Valley State University Gogebic Community College Mott Community College Michigan Technological University Grand Rapids Community College Muskegon Community College Michigan State University Henry Ford Community College North Central Michigan College Oakland University Jackson Community College Northwestern Community College Northern Michigan University Kalamazoo Valley Community College Oakland Community College Saginaw Valley State University Kellogg Community College Schoolcraft College The University of Michigan-Ann Arbor Kirtland Community College Southwestern Michigan College University of Michigan-Dearborn Lake Michigan College St. Clair County Community College University of Michigan-Flint Lansing Community College Washtenaw Community College Wayne State University Macomb Community College West Shore Community College Western Michigan University 3 MME content area College courses used Mathematics College Algebra. Courses identified by 4-year universities. Reading Reading-heavy courses such as entry-level literature, history, philosophy, or psychology for 2-year universities. Courses identified by 4-year universities. Science Entry level biology, chemistry, physics, or geology for 2-year universities. Courses identified by 4-year universities. Social Studies Entry level history, geography, or economics for 2-year universities. 4 Grades were put on a numeric scale from 0-4 0=F 1=D 2=C 3=B 4=A Not used o AU, AWF, DR, R, RA, FR, T, TR, X Coded as 3.0 o P, CR Coded as 0.0 o IN, N, NC, NE, NS, W, WF, WP, WX, and U 5 Course Grade MME Score MME Content Area Sample Size Percent B or higher Mean SD Mean SD Math 6,286 47.0 2.49 1.18 1112.2 13.2 Reading 37,952 54.9 2.64 1.23 1117.2 24.6 Social Studies 39,721 54.4 2.63 1.22 1135.4 26.3 Science 15,608 50.0 2.54 1.19 1123.5 23.5 6 MME Content Area Course Type Number of Students Mathematics College Algebra 6567 Literature 456 American History 1731 Other History 3010 Psychology 16231 Sociology 8236 Political Science 6114 Philosophy 1869 Other 2517 Reading 7 MME Subject Area Science Social Studies Course Type Number of Students Biology/Life Science 8355 General Chemistry 5807 Physics 535 Other 1483 American History 1734 Other History 3006 Psychology 16230 Sociology 8231 Geography 612 Political Science 6108 Economics 3498 Other 2361 8 Students receiving an A Students receiving a B or better Students receiving a C or better Students receiving a B or better in 4-year universities Students receiving a B or better in 2-year institutions 9 Logistic Regression (LR) o Identify score that gives a 50% probability of achieving an A o Identify score that gives a 50% probability of achieving a B or better o Identify score that gives a 50% probability of achieving a C or better Signal Detection Theory (SDT) o Identify scores that maximize the proportion receiving consistent classifications from MME to college grades • i.e., both proficient/advanced and receiving a A/B/C or better • i.e., both not proficient/partially proficient and receiving a A-/B-/C- or worse o Equivalent to LR under mild monotonicity assumptions Selected SDT as the preferred method because of its purpose (maximizing consistent classification) 10 Where • success is obtaining an A/B/C or better • e is the base of the natural logarithm • β0 is the intercept of the logistic regression • β1 is the slope of the logistic regression • x is the MME score 11 Percent of Students Earning a B or Better Logistic Regression of Test Scores on College Grades (Using Simulated Data) 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 950 1000 1050 1100 Test Score 1150 1200 1250 12 Percent of Students Earning a B or Better Logistic Regression of Test Scores on College Grades (Using Simulated Data) 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 950 1000 1050 1100 Test Score 1150 1200 1250 13 Percent of Students Earning a B or Better Logistic Regression of Test Scores on College Grades (Using Simulated Data) 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 950 1000 1050 1100 Test Score 1150 1200 1250 14 Percent of Students Earning a B or Better Logistic Regression of Test Scores on College Grades (Using Simulated Data) 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 950 1000 1050 1100 Test Score 1150 1200 1250 15 Freshman Grade (known cut score) MME (unknown cut score) B- or Lower B or Higher College Ready Inconsistent Consistent Not College Ready Consistent Inconsistent Basic Idea Set the MME cut score to… Maximize the number of students in the Consistent cells Minimize the number of students in the Inconsistent cells Maximize consistent classification from MME to first-year college grades 16 MME Content Area Score 0 Known Cut Score B- or lower 0.2 0.4 0.6 B or higher1 0.8 1.2 Grade in First Related Freshman Credit Bearing Course 17 MME Content Area Score 82.8% consistent classification 0 Consistently Classified Inconsistently Classified Known Cut Score Unknown Cut Score B- or lower 0.2 0.4 0.6 B or higher1 0.8 1.2 Grade in First Related Freshman Credit Bearing Course 18 82.8% consistent classification MME Content Area Score Adjust the unknown cut score to maximize consistent classification 0 Consistently Classified Inconsistently Classified Known Cut Score Unknown Cut Score B- or lower 0.2 0.4 0.6 B or higher1 0.8 1.2 Grade in First Related Freshman Credit Bearing Course 19 MME Content Area Score 88.6% consistent classification 0 Consistently Classified Inconsistently Classified Known Cut Score Unknown Cut Score B- or lower 0.2 0.4 0.6 B or higher1 0.8 1.2 Grade in First Related Freshman Credit Bearing Course 20 MME Content Area Score 90.45% consistent classification 0 Consistently Classified Inconsistently Classified Known Cut Score Unknown Cut Score B- or lower 0.2 0.4 0.6 B or higher1 0.8 1.2 Grade in First Related Freshman Credit Bearing Course 21 Analyses treating grades of A as the success criterion produced unusable results (i.e., the highest possible MME scale scores Analyses treating grades of C as the success criterion produced unusable results (i.e., MME scale scores below chance level) Analyses treating 4-year and 2-year institutions did produce different cut scores, but they were within measurement error of each other Used analyses based on all institutions and grades of B or better to produce MME cut scores Used probability of success of 33% and 67% to set the “partially proficient” and “advanced” cut scores SDT and LR produced very similar results Used SDT because it was the preferred methodology 22 Content Area Classification Consistency Partially Proficient Cut Score Proficient Cut Score Advanced Cut Score Mathematics 65% 1093 1116 1138 Reading 63% 1081 1108 1141 Science 67% 1106 1126 1144 Social Studies 63% 1097 1129 1158 23 Identifying MEAP Cut Scores 24 Grade Cohort 3 4 5 6 7 8 9 10 11 12 13 1 - - - - - 05-06 06-07 07-08 08-09 09-10 10-11 2 - - - - 05-06 06-07 07-08 08-09 09-10 10-11 - 3 - - - 05-06 06-07 07-08 08-09 09-10 10-11 - - 4 - - 05-06 06-07 07-08 08-09 09-10 10-11 - - - 5 - 05-06 06-07 07-08 08-09 09-10 10-11 - - - - 6 05-06 06-07 07-08 08-09 09-10 10-11 - - - - - 7 06-07 07-08 08-09 09-10 10-11 - - - - - - 8 07-08 08-09 09-10 10-11 - - - - - - - 9 08-09 09-10 10-11 - - - - - - - - 10 09-10 10-11 - - - - - - - - - 25 Logistic Regression (LR) o Identify score that gives a 50% probability of achieving proficiency on a later- grade test (i.e., MME or MEAP) Signal Detection Theory (SDT) o Identify scores that maximize the proportion receiving consistent classifications from one grade to a later grade • i.e., proficient/advanced on both tests • i.e., not proficient/partially proficient on both tests o Equivalent to LR under mild monotonicity assumptions Equipercentile Cohort Matching (ECM) o Identify scores that give the same percentage of students proficient/advanced on both tests Selected SDT as the preferred method because of its purpose (maximizing consistent classification) However, SDT and LR are susceptible to regression away from the mean 26 Same as for identifying MME cut scores Criterion for success is proficiency on a later grade test rather than getting a B or better in a related college course 27 Signal Detection Method (Simulated Data) Grade 8 Score Each dot represents a plot of test scores in grade 8 and grade 11 for a single student Known Cut Score Unknown Cut Score Grade 11 Score 28 Signal Detection Method (Simulated Data) Grade 11: Proficient Grade 8 Score Grade 11: Not proficient Known Cut Score Unknown Cut Score Grade 11 Score 29 Signal Detection Method (Simulated Data) Grade 8: Proficient Grade 11: Proficient Grade 8 Score Grade 8: Proficient Grade 11: Not proficient Known Cut Score Unknown Cut Score Grade 8: Not proficient Grade 11: Not proficient Grade 8: Not Proficient Grade 11: Proficient Grade 11 Score 30 Signal Detection Method (Simulated Data) Grade 8 Score Known cut score stays where it is Known Cut Score Unknown Cut Score Grade 11 Score 31 Grade 8 Score Signal Detection Method (Simulated Data) Known Cut Score Unknown Cut Score Move the unknown cut score up or down to maximize the same classifications in both grades Grade 11 Score 32 The more links in the chain, the greater the effects of regression Original plan for Math and Reading o Link grade 11 MME to college grades o Link grade 8 MEAP to grade 11 MME o Link grade 7 MEAP to grade 8 MEAP o Link grade 6 MEAP to grade 7 MEAP o Link grade 5 MEAP to grade 6 MEAP o Link grade 4 MEAP to grade 5 MEAP o Link grade 3 MEAP to grade 4 MEAP Original plan results in 7 links by the time the grade 3 cut is set Original plan results in inflated cut scores in lower grades 33 Revised plan for Math and Reading For Grade 3, 4, 5, 6 o Link grade 11 MME to college grades o Link grade 7 MEAP to grade 11 MME o Link grade 3, 4, 5, or 6 MEAP to grade 7 MME For Grade 7, 8 o Link grade 11 MME to college grades o Link grade 7 or 8 MEAP to grade 11 MME Results in a maximum of three links for any one grade 34 No evidence of regression away from the mean in identifying MEAP “proficient” cut scores o Looking for a consistently lower percentage of students proficient as one goes down in grades o Used SDT to identify MEAP “proficient” cut scores Evidence of regression away from the mean in identifying MEAP “partially proficient” and “advanced” cut scores o Increasingly smaller percentages of students in the “Not proficient” and “Advanced” categories as one goes down in grade o Used ECM instead to identify MEAP “Not Proficient” and “Advanced” cut scores 35 No evidence of regression away from the mean in identifying MEAP “proficient” cut scores o Looking for a consistently lower percentage of students proficient as one goes down in grades o Used SDT to identify MEAP “proficient” cut scores Evidence of regression away from the mean in identifying MEAP “partially proficient” and “advanced” cut scores o Increasingly smaller percentages of students in the “Not proficient” and “Advanced” categories as one goes down in grade o Used ECM instead to identify MEAP “Not Proficient” and “Advanced” cut scores 36 Classification Consistency Rates for MEAP Cut Scores in Mathematics Grade Cut Score Partially Proficient Proficient Advanced 8 83% 86% 95% 7 81% 84% 95% 6 82% 83% 96% 5 81% 84% 95% 4 80% 82% 94% 3 77% 80% 95% 37 Classification Consistency Rates for MEAP Cut Scores in Reading Grade Cut Score Partially Proficient Proficient Advanced 8 83% 78% 87% 7 86% 76% 85% 6 85% 74% 83% 5 88% 75% 84% 4 80% 82% 94% 3 80% 72% 86% 38 Classification Consistency Rates for MEAP Cut Scores in Science Grade Cut Score Partially Proficient Proficient Advanced 8 80% 84% 92% 5 76% 82% 92% 39 Classification Consistency Rates for MEAP Cut Scores in Science Grade Cut Score Partially Proficient Proficient Advanced 9 85% 81% 91% 6 81% 77% 91% 40 Creating Mini-Cuts for PLC Calculations in Reading and Mathematics 41 100 90 Conditional Standard Error of Measurement 80 70 60 50 40 30 20 10 0 205 255 305 Grade 3 Mathematics Scale Score 355 405 42 Conditional Standard Error of Measurement 70 60 50 205 255 305 Grade 3 Mathematics Scale Score 355 Advanced (A) Proficient (P) Partially Proficient (PP) Not Proficient (NP) 100 90 80 40 30 20 10 0 405 43 100 90 Conditional Standard Error of Measurement 80 70 60 50 40 30 20 10 0 205 255 305 Grade 3 Mathematics Scale Score 355 405 44 Conditional Standard Error of Measurement 70 60 205 255 305 Grade 3 Mathematics Scale Score 355 A-Mid P-High P-Mid P-Low PP-High PP-Low NP-High NP-Mid NP-Low 100 90 80 50 40 30 20 10 0 405 45 Year X Grade Y MEAP Performance Level Low Not Mid Proficient High Partially Low Proficient High Low Proficient Mid High Advanced Mid Year X+1 Grade Y+1 MEAP Performance Level Not Partially Proficient Proficient Proficient Adv Low Mid High Low High Low Mid High Mid M I I SI SI SI SI SI SI D M I I SI SI SI SI SI D D M I I SI SI SI SI SD D D M I I SI SI SI SD SD D D M I I SI SI SD SD SD D D M I I SI SD SD SD SD D D M I I SD SD SD SD SD D D M I SD SD SD SD SD SD D D M 46 New Versus Old Cut Scores 47 Percent Proficient Mathematics, Grade 11 100 90 80 70 60 50 40 30 20 10 0 Old Cut Scores New Cut Scores 07-08 08-09 09-10 10-11 School Year 48 Percent Proficient Mathematics, Grade 8 100 90 80 70 60 50 40 30 20 10 0 Old Cut Scores New Cut Scores 07-08 08-09 09-10 10-11 School Year 49 Percent Proficient Mathematics, Grade 7 100 90 80 70 60 50 40 30 20 10 0 Old Cut Scores New Cut Scores 07-08 08-09 09-10 10-11 School Year 50 Percent Proficient Mathematics, Grade 6 100 90 80 70 60 50 40 30 20 10 0 Old Cut Scores New Cut Scores 07-08 08-09 09-10 10-11 School Year 51 Percent Proficient Mathematics, Grade 5 100 90 80 70 60 50 40 30 20 10 0 Old Cut Scores New Cut Scores 07-08 08-09 09-10 10-11 School Year 52 Percent Proficient Mathematics, Grade 4 100 90 80 70 60 50 40 30 20 10 0 Old Cut Scores New Cut Scores 07-08 08-09 09-10 10-11 School Year 53 Percent Proficient Mathematics, Grade 3 100 90 80 70 60 50 40 30 20 10 0 Old Cut Scores New Cut Scores 07-08 08-09 09-10 10-11 School Year 54 New Versus Old Cut Scores 55 Percent Proficient Reading, Grade 11 100 90 80 70 60 50 40 30 20 10 0 Old Cut Scores New Cut Scores 07-08 08-09 09-10 10-11 School Year 56 Percent Proficient Reading, Grade 8 100 90 80 70 60 50 40 30 20 10 0 Old Cut Scores New Cut Scores 07-08 08-09 09-10 10-11 School Year 57 Percent Proficient Reading, Grade 7 100 90 80 70 60 50 40 30 20 10 0 Old Cut Scores New Cut Scores 07-08 08-09 09-10 10-11 School Year 58 Percent Proficient Reading, Grade 6 100 90 80 70 60 50 40 30 20 10 0 Old Cut Scores New Cut Scores 07-08 08-09 09-10 10-11 School Year 59 Percent Proficient Reading, Grade 5 100 90 80 70 60 50 40 30 20 10 0 Old Cut Scores New Cut Scores 07-08 08-09 09-10 10-11 School Year 60 Percent Proficient Reading, Grade 4 100 90 80 70 60 50 40 30 20 10 0 Old Cut Scores New Cut Scores 07-08 08-09 09-10 10-11 School Year 61 Percent Proficient Reading, Grade 3 100 90 80 70 60 50 40 30 20 10 0 Old Cut Scores New Cut Scores 07-08 08-09 09-10 10-11 School Year 62 New Versus Old Cut Scores 63 Percent Proficient Science, Grade 11 100 90 80 70 60 50 40 30 20 10 0 Old Cut Scores New Cut Scores 07-08 08-09 09-10 10-11 School Year 64 Percent Proficient Science, Grade 8 100 90 80 70 60 50 40 30 20 10 0 Old Cut Scores New Cut Scores 07-08 08-09 09-10 10-11 School Year 65 Percent Proficient Science, Grade 5 100 90 80 70 60 50 40 30 20 10 0 Old Cut Scores New Cut Scores 07-08 08-09 09-10 10-11 School Year 66 New Versus Old Cut Scores 67 Percent Proficient Social Studies, Grade 11 100 90 80 70 60 50 40 30 20 10 0 Old Cut Scores New Cut Scores 07-08 08-09 09-10 School Year 10-11 68 Percent Proficient Social Studies, Grade 9 100 90 80 70 60 50 40 30 20 10 0 Old Cut Scores New Cut Scores 07-08 08-09 09-10 10-11 School Year 69 Percent Proficient Social Studies, Grade 6 100 90 80 70 60 50 40 30 20 10 0 Old Cut Scores New Cut Scores 07-08 08-09 09-10 10-11 School Year 70 Joseph A. Martineau o Executive Director o Bureau of Assessment & Accountability o Michigan Department of Education o martineauj@michigan.gov o 517-241-4710 71