View
214
Download
1
Tags:
Embed Size (px)
Citation preview
[RESAMPLED RANGE OF WITTY TITLES]Understanding and Using the NRC Assessment of Doctorate Programs
Lydia Snover, Greg Harris & Scott BargeOffice of the Provost, Institutional Research
Massachusetts Institute of Technology • 2 Feb 2010
Overview1. Background & Context
2. Approaches to Ranking
3. The NRC Model: A Modified Hybrid
4. Presenting & Using the Results
OV
ER
VIE
W
2*NB: All figures/data in this presentation are used for illustrative purposes only and do not represent a known institution.
Participating MIT ProgramsAeronautics and Astronautics Computer Science
Applied Biosciences Economics
Applied Mathematics Electrical and Computer Engineering
Astrophysics and Astronomy and Planetary Science
Geology and Geochemistry and Geophysics
Atmospheres, Oceans & Climate including MIT/WHOI Oceanography
History, Theory and Criticism
Biological Engineering, Health Science and Technology
Linguistics
Biology/Biochemistry and Biophysics Material Sciences and Engineering
Biology/Cell and Developmental Mathematics
Biology/Genetics and Genomics Mechanical Engineering
Chemical Engineering Neuroscience
Chemistry Operations Research
Civil and Environmental Engineering Philosophy
Cognitive Science Physics
Computer Engineering Political Science 4
INT
RO
DU
CT
ION
How do we measure program quality?• Use INDICATORS (“countable” information)
to compute a rating– Number of publications
– Funded research per faculty member
– Etc.,
• Try to quantify more subjective measures through an overall PERCEPTION-BASED RATING– Reputation
– “Creative blending of interdisciplinary perspectives”
AP
PR
OA
CH
ES
TO
RA
NK
ING
S
6
So how does NRC blend the two?The NRC used a modified hybrid of the two basic approaches:•In total, a 4-step process, indicator based, by field
•Process results in 2 sets of indicator weights developed through faculty surveys:
– “Bottom up” –importance of indicators
– “Top-down” – perception-based ratings of a sample of programs
•Multiple iterations (re-sampling) to model “the variability in ratings by peer raters.” *
TH
E N
RC
AP
PR
OA
CH
8*For more information on the rationale for re-sampling, see pp. 14-15 of the NRC Methodology Report
So how does NRC blend the two?STEP 1: Gather raw data from institutions, faculty & external sources on programs. Random University (RU) submitted data for its participating doctoral programs.
TH
E N
RC
AP
PR
OA
CH
9
RU PhysicsIndicator Value
# publications/fac 1.07
# citations/article 1.17
Median GRE 746
Gender diversity 44%
Time to degree 5.67
…
RU ChemEIndicator Value
# publications/fac 1.07
# citations/article 1.17
Median GRE 746
Gender diversity 44%
Time to degree 5.67
…
RU EconomicsIndicator Value
# publications/fac 1.07
# citations/article 1.17
Median GRE 746
Gender diversity 44%
Time to degree 5.67
…
NRCNRC
So how does NRC blend the two?STEP 2: Use faculty input to develop weights:– Method 1: Direct prioritization of indicators--
“What characteristics (indicators) are important to program quality in your field?”
TH
E N
RC
AP
PR
OA
CH
10
Program Faculty Quality Most Impt Indicator (Mark 4)
Top 2 Indicators
Number of publications per faculty member
Number of citations per publication
Racial/ethnic diversity of the student population
Avg. # of Ph.D.s granted over last 4 years
Gender diversity of program faculty
… … …
Direct Weights
Indicator 1= 0.2
Indicator 2= 0.0
Indicator 3= 0.1
Indicator 4= 0.1
Indicator 5= 0.2
…
Calculations
So how does NRC blend the two?STEP 2: Use faculty input to develop weights:– Method 2: A sample of faculty each rate a sample of 15
programs from which indicator weights are derived.
TH
E N
RC
AP
PR
OA
CH
11
Program #2: Yale University Economics
Some Facts About the Program:# of Ph.D.s 2001-2006: _____ Faculty % Female: _____Median Time to Degree: _____ Current Faculty List, etc.
Program #2: Yale University EconomicsProgram #2: Ivy University Economics
Program #1: Land Grant University Economics
Some Facts About the Program:# of Ph.D.s 2001-2006: XX Faculty % Female: YY%Median Time to Degree: Z.Z Current Faculty List, etc.
On a scale from 1 to 3, indicate your familiarity with this program?___ 1 (Little or none)___ 2 (Some)___ 3 (Considerable)
On a scale from 1 to 6, how would you rate this program?___ 1 (Not adequate for doc educ.)___ 2 (Marginal)___ 3 (Adequate___ 4 (Good)___ 5 (Strong)___ 6 (Distinguished)___ 9 (Don’t know well enough)
Regression-based
Weights
Ind. 1= 0.3
Ind. 2= 0.04
Ind. 3= 0.2
Ind. 4= 0.15
…
PrincipleComponents& Regression
So how does NRC blend the two?STEP 3: Combine both sets of indicator weights and apply them to the raw data:
TH
E N
RC
AP
PR
OA
CH
12
Direct Weights
Ind. 1= 0.3
…
Regression-based
Weights
Ind. 1= 0.2
…
Combined Weights
Ind. 1= 0.25
…
DATAIndicatorValue
# publications/fac 1.07
# citations/article 1.17
Median GRE 746
Gender diversity 44%
Time to degree 5.67
…
X= Rating
RANKEDLIST
1. Ivy Univ (98)
2. Random Univ (94)
3. Private Univ (91)
4. Land Grant U (88)
5. Univ of State (87)
…
So how does NRC blend the two?STEP 4: Repeat steps 500 times for each field
TH
E N
RC
AP
PR
OA
CH
13
A) Randomly draw ½ of faculty
“important characteristics”
surveys
C) Randomly draw ½ of faculty
program rating surveys
B) Calculate “direct” weights
D) Compute “regression-
based” weights
E) Combine weights
F) Repeat (A) – (E) 500 times to develop 500 sets of weights for each field
G) Randomly perturb institutions’ program data 500 times*
H) Use each pair of iterations (1 perturbation of data (G) + 1 set of weights (F)) to rate programs and
prepare 500 ranked lists
I) Toss out the lowest 125 and highest 125 rankings for each
program and present the remaining range of rankings
*For more information on the perturbation of program data, see pp. 50-1 in the NRC Methodology Report
What are the indicators?Program Faculty Quality Student Characteristics Program Characteristics
# of publications per faculty member
Median GRE of entering students
Avg. # Ph.D.s granted in past 5 years
# of citations per faculty member
% students receiving full financial support
% entering students who complete
Receipt of extramural grants for research
% students with portable fellowships
Time to degree
Involvement in interdisciplinary work
Racial/ethnic diversity of student population
Placement of students after grad
Racial/ethnic diversity of the program faculty
Gender diversity of student population
% students with individual work space
Gender diversity of the program faculty
High % of international students
% of health insurance premiums covered
Reception of peers of a faculty member’s work as measured by honors/awards
# of student support activities provided
15
RE
SU
LT
S
What will the results look like?• TABLE 1: Program values for each indicator plus
overall summary statistics for the field
RE
SU
LT
S
16
RU Econ All Economics Programs (n=117)
Indicator Value Min 25th %tile
75th %tile
Max Std. Dev.
# publications/fac 1.07 .049 .369 .655 1.257 .246
# citations/article 1.17 .153 .684 1.771 5.485 1.002
Median GRE 746 353 740 790 800 55
% female students 44% 0% 28.6% 42.9% 76.9% 12%
% female faculty 12.5% 0% 10.5% 21.1% 66.7% 9.9%
Time to degree 5.67 3 5 6 8 .8
…
What will the results look like?• TABLE 2: Indicators and indicator weights – one
standard deviation above and below the mean of the 500 weights produced for each indicator through the iterative process (and a locally calculated mean)
RE
SU
LT
S
17
Indicator Minus 1 SD Plus 1 SD Calculated Mean
# publications/fac 0.130 0.134 0.132
# citations/article 0.294 0.267 0.2805
Median GRE 0.091 0.089 0.09
% female students -0.029 -0.043 -0.036
% female faculty n.s.* n.s.*
Time to degree -0.026 -0.031 -0.0285
…
*n.s. in a cell means the coefficient was not significantly different from 0 at the p=.05 level.
What will the results look like?• TABLE 3: Range of rankings for RU’s Economics
program alongside other programs, overall and dimensional rankings
RE
SU
LT
S
18
Institution Overall Research Activity
Diversity of Acad Environ.
Student Supp/Outcomes
25th %tile
75th %tile
25th 75th 25th 75th 25th 75th
Ivy Univ 30 36 31 32 37 41 28 31
Univ of State 45 54 40 42 42 50 45 46
Random Univ 45 56 38 42 47 51 43 47
Private Univ 48 57 41 42 40 47 45 49
Land Grant U 55 63 59 64 48 50 54 61
Total # of ranked programs = 117
What will the results look like?• TABLE 4: Range of rankings for all RU’s programs
RE
SU
LT
S
19
Program Overall Research Activity
Diversity of Acad Environ.
Student Supp/Out
comes
1995 NRC
Ranking
2009 USNWR Ranking
25th 75th 25th 75th 25th 75th 25th 75th
Linguistics 45 56 … 40 38Material Sciences and Engineering
25 26 24 24
Mathematics 21 23 23 25Mechanical Engineering
32 36 33 33
Neuroscience 34 35 34 35Operations Research
54 56 56 53
Philosophy 43 44 … 44 43