Elementary Statistics: STEM vs. nonSTEM Dave Usinski: [email protected] Lynn Meslinsky: [email protected]
Feb 23, 2016
Elementary Statistics:STEM vs. nonSTEM
Dave Usinski: [email protected] Meslinsky: [email protected]
Our Question
What are the best predictors of success in Statistics and how can we use these to determine whether this course should be STEM or nonSTEM?
Overview
• Revisit MT112/MT116 non-STEM Pathway– Has it been successful?
• Incorporate Placement Data• Kendall’s Tau B (rows = columns) and C
– Contingency table rank correlation coefficients between numeric vs ordinal and ordinal vs ordinal
• Binary Logistic Regression Analysis– Predicting future course success
We looked at our other non-STEM courses
How did lessening the algebra prerequisite requirements affect the course success rates?
Overall College Fall Headcount, Number of Students Passing MT112/116 in Fall,
Number of Students Passing MT112/116 in entire Year
02/FA 03/FA 04/FA 05/FA 06/FA 07/FA 08/FA 09/FA 10/FA 11/FA 12/FA0
2000
4000
6000
8000
10000
12000
14000
16000
Student Totals
Fall HeadcountFall passing MT112/MT116Year passing MT112/MT116St
uden
ts
Same as Previous except as Percent Change from Fall 2002
02/FA 03/FA 04/FA 05/FA 06/FA 07/FA 08/FA 09/FA 10/FA 11/FA 12/FA-20.00%
0.00%
20.00%
40.00%
60.00%
80.00%
100.00%
120.00%
Percent Change based on Fall 2002
% change Fall Headcount% change Fall Passing MT112/MT116% change Yearly passing MT112/MT116
Perc
ent C
hang
e
We looked at the last course taken prior to taking Statistics
Which courses had the best predictive value in terms of success in Statistics?
We also compared placement results in reading to placement
results in algebra/math.
Which Accuplacer or SAT result had a higher correlation to success with Statistics?
Correlation CoefficientDescribing Strength of Association
Characterizations
>.5 high association
.3 to .5 moderate association
.1 to .3 low association
0 to .1 little if any association
MT143 (Statistics) PredictorsPredictor Kendall’s Tau
B or CCorrelation Coefficient95% Confidence Interval
AccuPlacer Reading (55 to 110)
C n=953 0.135 (0.089 to 0.181)
AccuPlacer Algebra (55-110) C n=793 0.107 (0.059 to 0.156)SAT M (300-700) C n=2722 0.146 (0.120 to 0.172)SAT V (300-700) C n=2675 0.091 (0.064 to 0.119)MT013 (Elementary Algebra) B n=707 0.265 (0.206 to 0.325)MT112 (Survey of Math) B n=931 0.328 (0.278 to 0.378)MT125 (College Alg. & Trig I) B n=2156 0.380 (0.349 to 0.412)MT126 (College Alg. & Trig II)
B n=690 0.360 (0.305 to 0.415)
MT175 (Survey of Calculus) B n=735 0.340 (0.286 to 0.395)
MT175
MT126
MT125
MT112
MT013
SAT M
SAT V
AP Alg
AP Reading
0.450.400.350.300.250.200.150.100.05Data
Boxplot of AP Reading, AP Alg, SAT V, SAT M, MT013, MT112, MT125, ...
43210
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
grade
Y-Da
ta
0.6
>=B- Probability>=C+ Probability>=C Probability>=C- Probability
Variable
MT013 grade vs MT143 grade
43210
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
grade
Y-Da
ta
0.6
>=B- Prob>=C+ Prob>=C Prob>=C- Prob
Variable
MT112 grade vs MT143 grade
Conclusion
There is little difference in the predictive value of Developmental Algebra vs. Liberal Arts Math in determining a grade in Statistics.
120100806040200
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
score AP Read
Y-Da
ta
0.6
0.75
>=B- Prob APREAD>=C+ Prob APREAD>=C Prob APREAD>=C- Prob APREAD
Variable
AccuPlacer Read vs MT143 Prob
120100806040200
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
score AP Alg
Y-Da
ta
0.6
0.75
>=B- Prob APALG>=C+ Prob APALG>=C Prob APALG>=C- Prob APALG
Variable
AccuPlacer Algebra vs MT143 Prob
Conclusion
An Accuplacer reading score of about 63 is needed to have a 60% chance of C or better in Statistics while an Accuplacer Algebra Score of about 30 will do the same thing.
800700600500400300200
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
score SAT M
Y-Da
ta
0.6
0.75
>=B- Prob SATM>=C+ Prob SATM>=C Prob SATM>=C- Prob SATM
Variable
SAT MATH vs MT143 Prob
800700600500400300200
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
score SAT E
Y-Da
ta
0.6
0.75
>=B- Prob SATE>=C+ Prob SATE>=C Prob SATE>=C- Prob SATE
Variable
SAT Eng vs MT143 Prob
Conclusion
SAT score showed the opposite! A higher SAT Math, 270, was needed to show 60% chance of C or better in Statistics while only a SAT Verbal had no impact.
Questions?
• Dave Usinski: [email protected]• Lynn Meslinsky: [email protected]