Using%20Demographic%20Data%20to%20Predict%20Students

1
Using Demographic Data to Predict Students
Achievement in DSPM Courses

• Daryl Stephens, ETSU
• stephen_at_etsu.edu
• TNADE
• October 27, 2005

2

• Students take developmental studies program (DSP)
  courses for many reasons
• Forgot material learned in high school by the
  time they took entrance tests
• Long gap between high school and matriculation
• Didnt take high school seriously
• Didnt consider going to college until later
• First generation college student
• (Salter Noblett, 1994)

3

• Most students (90) at ETSU take MATH 1530,
  Probability and Statistics, as their mathematics
  course to satisfy core curriculum requirements.
• In previous years, this course had a high failure
  rate.

4
The Problem

• Placement procedures for TBR schools have
  changed.
• Pre-2000
• ACT lt 19 or age gt 22 ? take AAPP
• 2000-2002 COMPASS replaces AAPP
• 2002 ACT score alone determines placement for
  students with score lt 3 years old COMPASS alone
  for others

5
The Problem

• State funding is static.
• Placement decisions based on one test.
• Enrollment caps may happen in the future.
• Is there a way to augment the placement process
  by predicting a students chance of success or
  failure in developmental or core math classes?

6
Purpose

• Develop models to predict success of students in
• DSPM 0800 (Elementary Algebra)
• DSPM 0850 (Intermediate Algebra)
• MATH 1530 (Probability and Statistics)
• Use multiple regression to develop the models
  using readily obtainable information

7
Importance

• MATH 1530 traditionally had high failure rate
• Developmental math students are at a greater risk
  of failure and dropping out
• Specific developmental studies program advisors
  done away with in budget crunch of 2003

8
Importance

• Previous similar studies on developmental
  students used additional instruments which cost
  money.
• Very few universities require probability and
  statistics for math credit for graduation. Most
  similar studies on core math deal with college
  algebra, precalculus, or math survey courses.

9
Assumptions

• Prediction is possible
• Self-reported data are correct
• Information in SIS is correct

10
Limitations

• Different demographics and course requirements
  from other institutions, so work only applies to
  ETSU
• Data only collected for fall spring and summer
  grades are probably different

11
Definitions

• DSPM 0800 Elementary algebra
• Arithmetic review, algebraic representations,
  linear equations in one and two variables
• DSPM 0850 Intermediate algebra
• Exponents, polynomials, factoring,
• MATH 1530 Probability and statistics
• Stat mansion and stat cave

12
Definitions

• COMPASS (Computerized Adaptive Placement
  Assessment and Support System
• Prealgebra and algebra sections
• Reading and writing sections
• Replaced AAPP

13
Previous Research
14
Developmental Studies History

• Special programs at Harvard in 17th century
• Preparatory department at University of
  Wisconsin, 1849
• Morrill Acts, late 19th century, establish land
  grant colleges and extend access to higher
  education to more people

15
Developmental Studies History

• Preparatory departments widespread in early 20th
  century (350 in 1915)
• GI Bill of Rights brings in veterans with needs
  for auxiliary services
• 1960s-70s Increase of women, students of color,
  first-generation students, students with learning
  disabilities open-access community colleges
  established

16
Tennessee DSP History

• TBR establishes formal developmental studies
  program in 1984.
• Defining Our Future (2001) operate more
  efficiently and with more limited resources
• Move 0700-level courses to community colleges

17
Related Research

• Developmental Math
• Core Math (almost nothing on PS)
• What factors are related to success in the
  courses? Three broad categories
• Academic
• Demographic
• Affective

18
What variables predict course success in dev.
math?

• High school GPA, at least for traditional
  students
• Scores on the ACT or SAT sometimes
• Placement tests (e.g. COMPASS, ASSET, Accuplacer,
  CPT, PTT, CLAST) sometimes
• GED math scores as adjunct to other placement
  scores

19
Course success in dev. math

• Number of high school mathematics courses taken
  sometimes
• High school math GPA some courses
• College GPA
• Attendance
• Study habits
• Age or length of time since last math class

20
Course success in dev. math (contd)

• Gender (some studies)
• Race
• Math anxiety level not related to grade!
• Attitude
• Perception of success, engagement in class
• Paying attention and interacting with instructor

21
Success in core classes

• Very little research on success in classes like
  MATH 1530, so other courses examined
• ACT/SAT math scores (usually)
• COMPASS, TASP, local placement tests
• High school GPA for college algebra (multiple
  studies) and calculus (but not precalculus)

22
Success in core classes (contd.)

• HS math GPA mixed results
• HS percentile rank
• Number and difficulty of HS math classes taken
• Whether a math class taken senior year
• Students who didnt take intermediate algebra
  scored sig. higher in college alg.

23
Success in core classes (contd.)

• Age in some cases
• Time since last math course
• Gender? Yes in 3, no in 5 studies
• Full-time vs. part-time (1 study)
• Class meeting time (1 study)
• No difference in resident vs. commuter, campus
  activity

24
Success in core classes (contd.)

• Attitude
• Learning styles
• Self-concept

25
Research Questions
26
Relationship Questions

• Is there a relation between course grade and
• ACT (DSPP) math scores?
• ACT (DSPP) reading scores?
• COMPASS intermediate algebra scores?
• COMPASS reading scores?
• Number of college preparatory math classes taken
  in high school?
• High school GPA?

27
Regression QuestionsCan a regression equation be
found to predict final course grade based on
these items?

• From SIS
• ACT composite
• ACT math
• ACT reading
• ACT English
• High school GPA
• Age on first day of class

• From Survey
• Number of high school math classes taken
• Number of years since last HS math class

28
Regression QuestionsCan a regression equation be
found to predict final course grade based on
these items?

• From SIS
• COMPASS writing
• COMPASS reading
• COMPASS prealgebra
• COMPASS intermediate algebra
• Age on first day of class

• From Survey
• Number of high school math classes taken
• Number of years since last HS math class

29
Method

• Collect information about courses taken in high
  school and year of last high school math class
  from students via survey
• Obtain other information from SIS
• Use Pearson correlation for relationship
  questions
• Use stepwise multiple regression for grade
  prediction questions

30
Initial Placement

• ACT (DSPP) Math section
• lt 17 (SAT lt 440) DSPM 0800
• 18 (SAT 450) DSPM 0850
• gt19 (SAT gt 450) college level math
• COMPASS
• Prealgebra score lt 29 DSPM 0700
• Prealgebra 30-99 and algebra 20-27 DSPM 0800
• Algebra 28-49 DSPM 0850
• Algebra 50-99 college level math

31
Data considerations

• Not counted grades of W, WF, I
• FN grade not counted to be consistent with some
  other ETSU studies
• Online sections, RODP sections not included

32
Surveys returned

• DSPM 0800 149 / 304 (49)
• No night, off-campus, or ITV sections
• DSPM 0850 214 / 455 (47)
• Included Kingsport, night, ITV
• MATH 1530 631 / 1074 (59)

33
Results
34
Grade Distribution
Grade DSPM 0800 DSPM 0850 MATH 1530
A 35 75 70
A- 17 18 48
B 14 17 35
B 21 21 98
B- 8 11 36
C 9 13 34
C 13 25 89
C- 35
D 42
D 28
F 19 19 49
FN 10 8 58
I 0 1 1
W 3 6 5
WF 0 0 3
Total 149 214 631
Mean 2.785 2.991 2.419
SD 1.2993 1.2075 1.1526

• Table 3
• Grades of C-, D, and D are not allowed in
  developmental studies courses.
• Not used in answering the research questions.

35
ACT Math vs. Grade
ACT Math
DSPM 0800 DSPM 0850 MATH 1530
r 0.025 .322 .481
GPTS p 0.802 0 0
N 102 169 499
36
ACT Reading vs. Grade
ACT Reading ACT Reading
DSPM 0800 DSPM 0850 MATH 1530
r 0.047 0.164 0.164
GPTS p 0.642 0.033 0
N 102 169 497
37
COMPASS Int. Alg. vs. Grade
COMPASS Intermediate Algebra Score COMPASS Intermediate Algebra Score COMPASS Intermediate Algebra Score
DSPM 0800 DSPM 0850 MATH 1530
r 0.107 0.326 0.133
GPTS p 0.448 0.029 0.348
N 52 45 52
38
COMPASS Reading vs. Grade
COMPASS Reading Score COMPASS Reading Score COMPASS Reading Score
DSPM 0800 DSPM 0850 MATH 1530
r 0.222 0.206 0.118
GPTS p 0.194 0.323 0.513
N 36 25 33
39
High School Math Courses(Most Common)
Course 0800 0850 1530
Prealgebra 47 37 67
Foundations 2 16 26 28
Algebra 1 123 181 518
Algebra 2 111 182 556
Geometry 102 176 544
Precalculus 6 27 211
Trig/Adv. Alg. 4 27 133
Prob. Stat. 4 6 72
Calculus 2 4 88
N 149 214 633
40
Course Grade vs. Number of College Prep Classes
Taken
HSMATH (All students) HSMATH (All students) HSMATH (All students) HSMATH (All students)
DSPM 0800 DSPM 0850 MATH 1530
r -0.068 0.191 0.265
GPTS p 0.434 0.007 .000
N 135 198 564
Mean taken 2.52 3.11 3.73
SD taken 1.034 0.955 1.115
41
Course Grade vs. Number of College Prep Classes
Taken
HSMATH (Students with ACT Scores) HSMATH (Students with ACT Scores) HSMATH (Students with ACT Scores) HSMATH (Students with ACT Scores)
DSPM 0800 DSPM 0850 MATH 1530
r -0.237 0.193 0.298
GPTS p 0.017 0.012 .000
N 102 168 499
Mean taken 2.84 3.28 3.85
SD taken 1.2333 1.2 1.034
42
Course Grade vs. Number of College Prep Classes
Taken
HSMATH (Students with COMPASS scores) (Students with COMPASS scores) (Students with COMPASS scores) (Students with COMPASS scores)
DSPM 0800 DSPM 0850 MATH 1530
r -0.061 0.131 0.171
GPTS p 0.671 0.392 0.226
N 51 45 52
Mean taken 1.88 2.53 2.85
SD taken 1.103 1.254 1.036
43
HSGPA vs. Grade
HS Overall GPA HS Overall GPA HS Overall GPA
DSPM 0800 DSPM 0850 MATH 1530
r .389 .333 .445
GPTS p .000 .000 .000
N 105 176 500
44
Regression Using ACT

• DSPM 0800 (95 students)
• Model 1
• y  .589(HSMATH)  4.599 (p .001)
• Model 2
• y   .765(HSMATH)  1.009(HSOGPA)  2.298 (p
  lt .001)

45
Regression Using ACT

• DSPM 0850 (160 students)
• Model 1
• y   .364(ACTM)  3.238 (p lt .001)
• Model 2
• y .301(ACTM)  .662(HSOGPA)  4.111
• (p lt .001)

46
Regression Using ACT

• MATH 1530 (475 students)
• Four models (p lt .001 in each case r2 ranging
  from .233 to .353)
• y   .134(ACTM)  .521
• y   .092(ACTM)  .771(HSOGPA)  2.181
• y   .103(ACTM)  .910(HSOGPA) 
•  .108(AGE)   4.953
• y   .089(ACTM)  .855(HSOGPA)  
•   .105(AGE)  .025(ACTE) 4.9

47
Regression Using COMPASS

• DSPM 0800 (33 students) no equation possible
• DSPM 0850 (23 students with all sections)
• y .020 (COMPASS writing) 1.374
• (r2 .221, p .024)
• DSPM 0850 (44 students with math scores)
• y .023 (COMPASS arithmetic) 1.730
• (r2 .190, p .033)

48
Regression Using COMPASS

• MATH 1530
• 22 students with all COMPASS sections no
  equation possible
• 51 students with just math sections
• y .027(COMPASS arithmetic) .671
• (r2 .276, p lt .001)

49
Observations, Conclusions, and Recommendations
50
Observations 0800

• Elementary algebra only high school GPA showed a
  significant correlation with course grade math
  preparation in high school showed a little bit of
  predictive value.
• Finding agrees with Long (2003) ACT math score
  not significantly correlated with E.A. course
  grade

51
Observations 0800 (contd)

• Odd finding Why a negative correlation between
  HS math prep and grade in 0800?

52
Observations

• Intermediate algebra course grade correlated
  with ACT math and reading, COMPASS intermediate
  algebra and reading, high school math
  preparation, high school GPA.
• ACT math correlation agrees with Lott (1990)
• HS math preparation agrees with 4 studies

53
Observations

• Probability and statistics grade was
  significantly correlated with ACT math and
  reading and HSGPA, but regression used ACT math
  and English, age, and HSGPA. Only COMPASS
  arithmetic score was in regression for students
  without ACT scores.
• Findings in line with 7 other studies of other
  core college math courses.

54
Recommendations

• For students with borderline placement scores,
  use the regression equation to fine-tune the
  placement.
• If students regression equation predicts a
  failing grade, give extra attention to that
  student (watch attendance, recommend tutoring,
  etc.), or consider placement in a lower level
  class if possible.

55
Future Research

• Repeat study with data from spring and summer.
  Historically, grades in developmental math
  classes are significantly higher in summer and
  lower in spring. (Unknown if true for 1530)
• Repeat treating repeaters separately, or include
  number of previous attempts as a variable

56
Future Research

• Is grade in 1530 affected by other college-level
  or developmental math experience?
• Characteristics of repeating students (mostly
  qualitative research)
• What if core curriculum at ETSU changes allowing
  other math courses to count for graduation?
• How would raising cutoff scores affect grades?
  (TBR uses 19, ACT recommends 22)

57
Future Research

• Qualitative study (suggested by a student) What
  kind of guidance were students given in selecting
  high school math classes?
• Why a negative correlation between HS math prep
  and 0800 grade?

58
For more information

• Entire dissertation is available
  onlinehttp//faculty.etsu.edu/stephen/
• Will be online from UTK in Jan. or Feb., but will
  stay on my faculty web page unless theres a
  space crunch.
• This slide show is also on my faculty web page.
  Look for the conference presentations link.