Sie sind auf Seite 1von 20

MA 105 ANALYSIS

Ma 404

BRAZEAL, JACOB

Ma 404: Ma 105 Analysis Jacob Brazeal 2

Introduction
In this analysis of the data from Ma 105 students, we look at how performance in Ma 105 is
related to a number of other variables, such as GPA, ACT scores, and demographics. We also
consider how those other variables are correlated with each other, and study some interesting
subpopulations. The last section discusses the effects of talent vs. work ethic on grades, based on
a special scoring system used quantify those traits.

The data includes a large number of students who dropped the class. Quantifying the
performance (such as average grade) of a group of students is difficult when many did not
receive a grade. Because many of these students would presumably have failed if they had stayed
in the class, and because we want to represent the whole population in our analysis, students who
drop the class are assigned a grade of 0 (lower than 1, an F) and included in the analysis, except
where noted.

Abstract
Our findings include the following: while school type and gender do not predict performance,
GPA and ACT scores do, and taking Ma103 before Ma105 typically predicts poorer performance
in the latter class. Also, taking the class in the spring vs. the fall is associated with poorer
performance.
Also, the three grade quartiles correspond to three core groups of students whose performance
can be accurately predicted using a system that scores talent and work ethic. A full table and
discussion of these results is included.
Ma 404: Ma 105 Analysis Jacob Brazeal 3

Modeling Performance
Summary
Let us start by summarizing student performance.1 Of the 134 students, the average score is a
5.81 (C+), and the standard deviation is 3.89. The following is a histogram of scores:

Ma 105 Grades
30

25

20

15

10

0
Dr F D- D C- C C+ B- B B+ A- A

The range of possible grades is encoded as follows (also showing quartiles):

0 (Dropped)
1 F
2 D- Q1
3 D
4 C-
5 C
6 C+
7 B- Median
8 B
9 B+ Q3
10 A-
11 A

We will proceed to consider a number of relevant questions.

1
For the idea to include this section and the abstract, the author is indebted to the sample paper.
Ma 404: Ma 105 Analysis Jacob Brazeal 4

General characteristics

With 95% confidence how many students can we expect to pass Ma 105?
In the data set, 76.8% of students pass the class. Using a two tailed one population z-test with
equal variances, the data supports the conclusion that 69.65% to 83.95% of the class will pass
Ma105.

Significantly, this indicates that anywhere from 16% to 30% of students will drop or fail the
class. As the histogram above indicates, these two categories dwarf the rest of the grade curve.
The simplest way to increase the average grade would be to keep these students in the class.

With 95% confidence is there a different pass rate for students taking Ma105 for the
second time?
A chart may be helpful here to visualize the differences between the population of 127 students
taking the class for the first time and the 7 retaking it.

Comparison of Pass Rates (%)


120
100
80
60
40
20
0
Not retaking Retaking

Pass Did not pass

Using a two-tailed chi-square test with 1 degree of freedom, the data supports the conclusion that
pass rates for students taking the class for the first time and for the second time are independent,
with a p-value of 0.848. Specifically, the 95% confidence interval is (-2.7646%, 41.5468%).

That is, students who retake the class likely have much lower odds of passing than students who
are taking it for the first time. These numbers are the first indication that there are subpopulations
in the class who interact with the material in substantially different ways. This idea will be
developed more later on.
Ma 404: Ma 105 Analysis Jacob Brazeal 5

Effect of ACT Scores

What is the average Math ACT score of students who pass Ma 105 with 90% confidence?
What is the difference between the average Math ACT scores of students who pass and
who don't pass with 90% confidence?

Of the 103 students who passed, the average Math ACT score is 26 and the standard deviation is
3.09. Using the two-tailed single population mean z-test, the data supports the conclusion that the
average Math ACT score of students who pass Ma 105 is between 25.5 and 26.5.

Of the 31 students who did not pass, the average score is 23 and the standard deviation is 3.07.

Using the two-tailed two-population mean t-test, the difference between the math ACT scores of
those who pass and don’t pass is between 1.96 and 4.04.

That is, students who pass the class average somewhere between 2 and 4 points on the ACT
Math test over students who don’t pass. The Math ACT turns out to be one of the most powerful
predictors of student success in the class.

Ma 105 Grade vs. Math ACT


40

35

30

25

20 R² = 0.20667
15

10

0
0 2 4 6 8 10 12

Is the student's grade in Ma 105 consistent with their Math ACT score?

Using a linear regression, we find that the correlation coefficient between Math ACT and Ma
105 Grade is R2 = 0.21. This value is indicates a weak correlation.
Ma 404: Ma 105 Analysis Jacob Brazeal 6

Math ACT vs. Ma 105 Grade


40

35

30

25

20
R² = 0.20667
15

10

0
0 2 4 6 8 10 12

Is the student's grade in Ma 105 consistent with their Composite ACT score?

Using a linear regression, we find that the correlation coefficient between Math ACT and Ma
105 Grade is R2 = 0.21. This value is indicates a weak correlation.

Composite ACT Score vs. Ma 105 Grade


35

30

25

20
y = 0.2864x + 23.894
15 R² = 0.11933

10

0
0 2 4 6 8 10 12

Effect of GPA

Is the student’s grade in Ma 105 consistent with their GPA?

Using a linear regression, we find that the correlation between Ma 105 grade and GPA is R2 =
0.119. This value indicates a very weak correlation.
Ma 404: Ma 105 Analysis Jacob Brazeal 7

The line of best fit indicates that increasing class grade by 1 step (e.g., from B to B+)
corresponds with 0.13 GPA points on average. GPA by itself is not enough to accurately predict
class performance, but it does provide insight when combined with other variables like ACT
scores.

Ma 105 Grade vs. GPA


4.50
4.00
3.50
3.00
2.50
2.00 y = 0.1334x + 2.0258
R² = 0.52926
1.50
1.00
0.50
0.00
0 2 4 6 8 10 12

Is a student’s GPA consistent with his Composite ACT score?

The correlation coefficient is R2 = 0.15, which indicates that relationship is weak but consistent.

To be clear, a higher GPA is correlated with a better ACT score – it’s just that lots of students
have high ACT scores and low GPAs, as the chart below demonstrates.

GPA vs. Composite ACT scores


35

30

25

20

15 R² = 0.15477

10

0
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50
Ma 404: Ma 105 Analysis Jacob Brazeal 8

Effect of High School Background


Does one high school type better prepare students for Ma 105 than the others? Is grade
independent of high school type?

We must first remove the “unknown” high school type, which has produced only failing grades
in the class. That high school has poor results.
Type of School2 n Average Std. Deviation
Unknown 4 0.0 0
BJA 15 6.7 3.6
AHE 3 4.3 4.0
Christian 75 6.2 3.9
Home School 24 5.3 3.6
Public 13 6.0 3.7

Using a single-factor ANOVA, the data does not support the conclusion that the mean grade
differs between the other groups (p=0.747, F=0.485, α=0.05). All (named) high school choices,
remarkably, produce roughly the same quality of math student.

To examine whether grade is independent of school type, we examine the distribution of grades
from different schools. To accommodate the small number of AHE students, we grouped them
with Home Schoolers; we grouped D-’s with F’s and withdrawals.

Using a Chi-squared test with 9 degrees of freedom, the data do not support the conclusion that
grades are independent of high school type (χ2=9.492, p=0.393, α=0.05).

No school type is especially good or bad at preparing students for math classes. This may be
encouraging to homeschooling parents with no mathematics education. It also suggests that
students do not struggle because of a particular teacher in high school.

Effect of Schedule & Course Sequence


Do students who take Ma 103 first do better in Ma 105 than students who did not take Ma
103? Can we assume that the variability of students who have taken Ma 103 first and those
who haven't is the same?

Students who have already taken Ma 103 (n = 55) average 4.5 (C) in Ma 105, with a standard
deviation of 3.74. Students who have not taken Ma 103 (n=79) average 6.7 (B-) in Ma 105, with
a standard deviation of 3.76.

2
The author is grateful to the sample paper for the idea to include descriptive stats here, and also
the idea to group sections by “Gender”, “ACT”, and so on.
Ma 404: Ma 105 Analysis Jacob Brazeal 9

Using a two-population, two-tailed mean t test with equal variance, the data support the
conclusion that students who have taken 103 prior to 105 score worse by between 0.91 and 3.51
letter-grade points (p=0.001, t=3.36, α=0.05).

This result might seem surprising; is Ma 103 harming students? In reality, the better explanation
is sample bias – weaker students are more likely to be placed into the remedial class. Just as
students who retake Ma 105 continue to struggle, students who have previously taken Ma 103
also continue to struggle.

The data supports the conclusion that the variances of the two groups is the same, using a two-
population, two-tailed F-test (p=0.494, F=0.99, α=0.05). This indicates that, despite a large
reduction in absolute performance, roughly the same grade curve exists in the Ma 103 cohort as
in the other. The overall aptitude is simply lower.

Do students who took Ma 103 get the same (or similar) grade in Ma 105 as they did in Ma
103?

As mentioned earlier, students who have already taken Ma 103 (n = 55) average 4.5 (C) in Ma
105, with a standard deviation of 3.74. Those same students averaged 8.07 with a (B) in Ma 103,
with a standard deviation of 2.25.

This statistic is perhaps skewed by the fact that we are selecting for students who did not drop
Ma 103, while many of the students did drop Ma 105. If we compare only students who do not
drop Ma 105, we find that those students average an 8.75 (B+) in Ma 103 (standard deviation
1.94) and 6.2 (C+) in Ma 105 (standard deviation 2.94) – still a huge drop.

Furthermore, none of the 55 students improved their grade from Ma 103 to Ma 105, while only 6
(11%) maintained it.3

Using a one-population, two-tailed mean t test with equal variance, the data does not support the
conclusion that the mean change in grade is zero ((p=0.000, t=9.995, α=0.05).

The Ma 103 students do not perform similarly in Ma 105; none of them improve, and on average
they drop by about a letter grade.

Consider three possible categories for placement test results: didn't take it, placed into Ma
103, and placed into Ma 105. Is the average Ma 105 grade different for these three groups?

Students who did not take the test (n=99) have a mean grade of 6.1 (C+) and a standard deviation
of 3.79. Students who placed into Ma 103 (n=19) have a mean grade of 3.2 (D), and a standard

3
Furthermore, the average work ethic score (see next section) of these 5 students was in the 86th
percentile.
Ma 404: Ma 105 Analysis Jacob Brazeal 10

deviation of 3.32. Students who placed into Ma 105 (n=16) have a mean grade of 7.3 (B-) and a
standard deviation of 3.94.4

The following graph groups failures and withdrawals together.

Ma 105 Grades vs. Placement Result


0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
A B C D F

Did not take test Placed into 103 Placed into 105

Using single-factor ANOVA, the data supports the conclusion that the average Ma 105 grade is
not the same for the 3 groups (p=0.002, F=6.31, α=0.05). However, a two-population two-tailed
t-test with equal variances supports the conclusion that the average Ma 105 grade is the same for
the group that did not take the test and the group that placed into Ma 105 (p=0.233, t=1.20,
α=0.05).
Therefore, if the group that did not take the placement test is our control group, we can conclude
that placing into Ma 105 indicates normal aptitude whereas placing into Ma 103 indicates a
deficiency.

Do the students in the fall do better in Ma 105 than the students in the spring? Is there
more variability in the fall or in the spring?

The students in the fall (n=82) have a mean of 6.51 and a standard deviation of 3.84. The
students in the spring (n=48) have a mean 5.04 and a standard deviation of 3.71.

Using a two-population mean t-test with equal variances, the data support the conclusion that
students perform better in the fall by between 0.12 and 2.82 grade points (p=0.035, t=2.13,
α=0.05).

4
Placement test results were inferred as following: students with 0’s were assumed not to have
taken the test; students who took the test and started in Ma 103 were assumed to have been
placed there; and students who started in Ma105, haven taken the test, were assumed to have
been placed there.
Ma 404: Ma 105 Analysis Jacob Brazeal 11

Therefore, the data strongly indicates that students perform better in the fall. There are many
reasons that could be proposed for this, such as a difference in teaching ability, but additional
analysis will reveal that the difference is strongly linked to Math ACT score; weaker students
tend to take the class in the spring (very likely because they took the remedial Ma 103 in the
fall).

The variance observed in the fall was 14.747, and the variance observed in the spring was
13.743. We can compare the variability using a two population two-tailed F-test. The data do not
support the conclusion that either semester has greater variability (p=0.402, F=1.07, α=0.05).

The fact that the spread of the grades in the spring is almost exactly the same as the spread of the
grades in the fall, albeit centered at a different mean, helps explain the next result.

Is the grade distribution (# As, Bs, etc.) different in the fall and the spring?

The following chart compares grades between the fall and the spring.

Distribution of Grades By Semester


0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
A B C D F

Fall Spring

For reasons mentioned earlier, we group D’s and F’s in our analysis. Using a Chi-Squared
independence test with 4 degrees of freedom, the data do not support the conclusion that grades
are distributed differently in the fall and in the spring (χ2=4.224, p=0.238, α=0.05).

How does the mean grade of students who start in Ma 105 differ between the fall and the
spring?
Among the 72 students who did not take Ma 103 and started Ma 105 in the fall, the average
grade is 7.0 (B-) and the standard deviation is 3.57. Among the 6 students who started in Ma 105
but in the spring, the average grade is 4.7 (C) and the standard deviation is 4.89.

Using a two-sample, two-tailed mean t-test assuming equal variances, the data do not support a
difference in grade between the fall and the summer. The confidence interval ranges from a 5.4
decrease in the spring to an 0.8 increase (χ2=4.224, p=0.141, α=0.05).
Ma 404: Ma 105 Analysis Jacob Brazeal 12

Although it is surprising to see this group of students struggling in the spring, there are too few
of them to draw any conclusions.

Effect of Gender
Do the male students do better in Ma 105 than the female students?

Men (n=82) have a mean of 5.47 and a standard deviation of 3.94. Women (n=42) have a mean
6.57 and a standard deviation of 3.71.

Using a two-population mean t-test with equal variances, the data do not support the conclusion
that men perform better than women (or that women perform better than men); p=0.128, t=1.53,
α=0.05.

Although women do a shade better than men in the listed data, the difference is not statistically
significant. One explanation that we will discover for this slight edge is that women seem to have
a better work ethic than men.

Do men have a greater chance of passing Ma 105 than the women have?

Of the 92 men, 68 (74%) pass, while of the 42 women 36 (86%) pass.

Using a two-population two-tailed proportion z-test, the data do not support the conclusion that
either the men or the women have a high probability of passing (p=0.129, z=1.53, α=0.05).

Even though more women passed the class in our dataset, our 95% confidence intervals of pass
rates for these two classes overlap: women could be between 75.13% and 96.29%, and men
could be between 64.94 and 82.88%. These intervals could overlap.

Is the grade distribution (# As, Bs, etc.) independent of gender?

It is important to note that the grades do not follow a standard distribution, but most likely mixed
one, because of the large tail of students who drop the class (22.3%).5 Therefore, we do not
attempt to present a formula for the distribution now.

Men and women have the following quartiles:

Men Women
Q1 1 5

5
In fact, we believe that the students have 3 sub populations, representing students at different
levels of talent and work ethic, and each with its own distribution (a fuller development of this
idea is in the last section of this paper).
Ma 404: Ma 105 Analysis Jacob Brazeal 13

Median 5.5 7
Q3 9 10

Their respective distributions appear as follows:

Men (n=92) Women (n=42)


30 16
14
25
12
20
10
15 8
6
10
4
5
2
0 0
A B C D F Dr A B C D F Dr

We group the withdrawals and F’s together. Using a Chi-square independence test with 4
degrees of freedom, the data does not support the conclusion that men and women have a
different distribution of grades (χ2=3.161, p=0.531, α=0.05).

Therefore, although women have slightly better grades in our data, we will refrain from building
a narrative out of that difference; we have no reason to suppose that it is not statistical noise.

Correlation of Major
Do the “tech” majors and “non-tech” majors have equal representation in the class?

There are 96 tech majors and 38 non-tech majors. Using a 1 population two-tailed proportion z-
test, the data support the conclusion that the two majors are not equal represented (p=0.000,
z=5.00, α=0.05).

There are more tech majors.

Do students in “tech” majors do better in Ma 105 than students in “non-tech” majors?

Of the 38 non-tech majors, the mean score is 5.3 (C) and the standard deviation is 3.9. Of the 96
tech majors, the mean score is 6.0 (C+) and the standard deviation is 3.9.

Using a two-population two-tailed mean t-test with equal variances, the data does not support the
conclusion that students in tech majors perform better than students in non-tech majors (p=0.329,
t=0.980, α=0.05).
Ma 404: Ma 105 Analysis Jacob Brazeal 14

First with gender and now with major, we do not find that broad labels affect student
performance.
Ma 404: Ma 105 Analysis Jacob Brazeal 15

Work Ethic vs. Talent


Teachers often discuss “work ethic” and “talent,” and speculate how various combinations of
these traits might affect a student’s grade. While it seems obvious that diligent geniuses will
succeed, and lazy imbeciles will struggle, our intuition might fail to predict the performance of
other combinations of these traits:

Strong work ethic Weak work ethic


Strong aptitude for math Good performance in Ma105 ?
Weak aptitude for math ? Poor performance in Ma105

Therefore, we assigned each student a Work Ethic score and a Talent Score, and placed the
student in a corresponding bucket.6 Then we computed descriptive statistics for each bucket.

Method
A student’s Talent Score was his ACT math percentile vs. the class.7
A student’s Work Ethic score seeks to show how he performed relative to his natural ability. It is
the percentile difference of his GPA percentile and his ACT composite percentile. Thus, a
student with 75th percentile GPA and a 50th percentile ACT would receive a raw Work Ethic
score of 25, scaled to his percentile. 8
The two scores are not entirely independent (since the Math ACT affects the composite ACT); a
higher talent level will tend to depress one’s apparent work ethic.

Summary
To account for students who were average, we broke each score into three buckets: 0% to 33%
(low), 34% to 66% (average), and 67% to 99% (high).

6
These formulae may rightly meet with some criticism but they allow some analysis of the
subject. Also, as metrics formulated entirely independent of performance in the class, they are
remarkably predictive.
7
Since this class is frequently one of the first math classes students take after the ACT, their
ability on the ACT translates nicely to their ability in the class. Of course, “talent” is a
generalization – much of the ability involved here, if not most of it, is due to a good education, a
lot of practice, and a good environment. A student’s talent, as roughly indicated by this score,
can easily change.
8
The ACT is professionally designed to indicate a student’s college readiness. GPA is the best
measure of a student’s actual college performance. Therefore, by comparing a student’s
percentile readiness to percentile performance, it is logical to attribute any improvement (or
underperformance) to effort, or lack thereof. The big caveat is that students who scored near
perfect on the ACT cannot have a positive raw work ethic, since there is no way to be better at
school (relative to peers) than they were predicted to be. However, by gaining a good GPA – the
best, or near the best – the student can at least maintain an “average” work ethic.
Ma 404: Ma 105 Analysis Jacob Brazeal 16

Each cell in the following table contains the mean grade of students in the given bucket, along
with the corresponding letter grade (rounding). Each bucket contains roughly one-ninth of the
students.

Low Work Ethic Average Work Ethic High Work Ethic

Low Talent 2.6 (D) 1.4 (F) 5.0 (C+)


Average Talent 4.1 (C-) 5.8 (C+) 8.7 (B+)
High Talent 5.4 (C) 9.6 (A-)9 10.0 (A-)

With 90% confidence, what is the difference in mean grade between students in high and
low percentile Talent Scores?

The low-talent students (n=45) have an average grade of 2 (D-) and a standard deviation of 0.82.
The high-talent students (n=45) have an average grade of 8 (B) and a standard deviation of 0.79.
Using the two-tailed two population t-test with equal variances, the data supports the conclusion
that the difference in means between high-talent and low-talent students ranges from 5.48 to 6.04
(α=0.1). (This is roughly two letter grades in the given scale.)

Talent is so important. Another way to think about talent is a student’s knowledge of math prior
to the course; if students know a lot, they automatically do far better than students who don’t,
and those students face a nearly insurmountable obstacle.

With 95% confidence, what buckets of Work Ethic and Talent do not have statistically
significant differences in mean?

Using single-factor ANOVA, the data support that the following buckets are grouped together
(we have assigned names to these groups, for reasons that will become clear):

Categories Test Description of Group


Statistics
At-risk Low talent, Low work ethic p: 0.337 n: 23
Low talent, Avg work ethic F: 0.97 Average: 2.00 (D-)
α: 0.05 Standard Deviation: 2.90
Normal Low talent, High work ethic p: 0.529 n: 71
Avg talent, Low work ethic F: 0.74 Average: 5.07 (C)
Avg talent, Avg work ethic α: 0.05 Standard Deviation: 3.32
High talent, Low work ethic
Gifted Avg talent, High work ethic p: 0.280 n: 38
High talent, Avg work ethic F: 1.32 Average: 9.42 (B+)
High talent, High work ethic α: 0.05 Standard Deviation: 2.10
9
The author found that he resides in this cell of the grid. Remarkably, it may predict his final
grade in every class he attends this semester.
Ma 404: Ma 105 Analysis Jacob Brazeal 17

So, for instance, low talent / high work ethic students look remarkably like high talent / low work
ethic students. They both get roughly a C. Interestingly, these three averages correspond exactly
with Q1, the median, and Q3 of the grades. Students at different places on the grade curve have
different characteristics, and these are what they are.

With 95% confidence, is there any difference in mean talent score between the spring and
fall semesters? Any difference in mean work ethic score?

The mean talent score for the 84 students in the fall semester is 0.62 (standard deviation: 0.25).
The mean talent score for the 48 students in the spring semester is 0.29 (standard deviation:
0.22). Using the two-tailed two population t-test with equal variances, the data supports the
conclusion that the difference in means between the buckets is significant at the 5% level
(p=0.000, t=7.57, α=0.05).

The biggest takeaway of this study is that students do far better in the fall because they have
better talent scores (Math ACTs).

The mean work ethic score for the students in the fall semester is 0.47 (standard deviation: 0.28).
The mean work ethic score for the students in the spring semester is 0.58 (standard deviation:
0.30). Using the two-tailed two population t-test with equal variances, the data does not support
the conclusion that the difference in means between the buckets is significant at the 5% level
(p=0.098, t=1.67, α=0.05).

Meanwhile, students in the spring actually exhibit a slightly better work ethic, but not one that is
statistically significant. Note that the work ethic is not calculated from the students’ behavior in
class, only their tendencies prior to joining class. So, sadly, struggling students are very likely
accustomed working harder than “talented” ones – but they are in a deep rut.

With 95% confidence, is there any difference in mean talent score between men and
women? Any difference in mean work ethic score?

The mean work ethic score for the 92 men is 0.46. For the 42 women, it is 0.57. Using the two-
tailed two population t-test with equal variances, the data supports the conclusion that the
difference in mean work ethic score is not significant at the 5% level (p=0.075, t= -1.893,
α=0.05).

The women have an 11-point better work ethic than the men in the data. This difference is not
quite statistically significant, but it’s probably why women have slightly better scores than men
in the class.

The mean talent score for men and women is the same, 0.50. If we take away one thing from this
analysis, it might be that men and women are exactly as smart as each other (at least by one
rubric)!
Ma 404: Ma 105 Analysis Jacob Brazeal 18

Are the At-risk and Gifted groups distributed exponentially?

We graph the groups’ grade distributions first.

Group Grade Distribution


0.6

0.5

0.4

0.3

0.2

0.1

0
Dr F D- D C- C C+ B- B B+ A- A

At-risk Normal Gifted

This graph makes clear why we named the groups as we did. “At risk” students follow a
distribution severely skewed toward dropping or failing the class, with a slightly bulky tail
containing the 13% scoring a B- or better. “Gifted” students follow a similar distribution skewed
towards high grades. “Normal” students follow no particular distribution (perhaps uniform?)–
they achieve every grade possible.10 This graph also explains the high incidence of B’s in the
overall distribution: all three distributions maintain a presence there.

Using the Lilliefors test on the At-risk students, with the estimated exponential parameter 0.522,
the data does not support the conclusion that the distribution is exponential (p = 0.004, θ=0.522,
α=0.05).

Using the Lilliefors test on the Gifted students, with the estimated exponential parameter 0.723,
the data does not support the conclusion that the distribution is exponential (p = 0.001, θ=0.723,
α=0.05).

The distributions clearly are not exponential.

10
Except for D, which no student achieves. That grade may represent a boundary between the
students who understand the material and turn in assignments vs. students who cannot
understand or don’t try.
Ma 404: Ma 105 Analysis Jacob Brazeal 19

Conclusion
The best predictor of student success in Ma 105 is the student’s GPA, followed by their Math
ACT score. The reason that scores are better in the fall is almost certainly because the best math
students, as measured by Math ACT scores, take the course then. Students who take the class in
the spring have significantly worse performance. The descriptive stats showed, intriguingly, a
better work ethic among students who took the class in the spring, but this difference was not
significant at the 0.05 level. Speculation that the teachers are better in the fall is unfounded;
performance is well explained by out-of-class factors.

Work Ethic Score and Talent Score (both calculated independently of performance in the class)
can model some of the subtler differences in performance. For example, although women’s
scores were stronger than men’s scores, this difference was not significant at the 0.05 level. The
difference that did exist could be attributed to women’s 11-point lead in Work Ethic score; there
was no difference in Talent Score between the groups.

In addition, breaking the population into groups based on Talent Scores and Work Ethic Scores
revealed three unique subpopulations within the class: poor performers (those having a low work
ethic and not highly talented) average a D-; average performers (those with average talent and
work ethic, or a high talent/low work ethic, and vice versa) average a C; and high performers
(those with a high talent score or work ethic and an at least average score in the other metric)
average a B+. Fascinatingly, the average score of each group corresponds to one of the three
quartiles of the grades.

Students who start in Ma 103 typically have lower ACT scores and lower talent, and continue to
struggle in Ma 105. This profile of student typically achieves a C only with a high work ethic;
anything less and the student frequently fails the class. Also, we may retire the ancient question
“who does better, the lazy genius or the diligent jock?” In Ma 105, they perform the same.

Teachers should focus on improving students’ talent (that is, their experience, intuition, and
understanding). As was shown earlier, students with high talent average two letter grades better
than students with low talent, across all levels of effort; they can achieve a good grade more or
less at will, if they’re willing to work for it. Struggling students know nothing but the struggle.
We must help them bridge that gap.
Ma 404: Ma 105 Analysis Jacob Brazeal 20

Appendix: Technology
For this assignment, the main tools used were R, Microsoft® Excel, and some additional online
utilities in the analysis problems.

R was primarily used to construct graphs and perform some simple analysis, as well as the
Lilliefors tests. The latter tests were the most distinctive task that R performs, since no resource
provided a method to calculate it adequately without a statistics package like R or SPSS.
Microsoft® Excel was used to filter data, to conduct additional descriptive analysis tests on
subsets of it, to perform F-Tests and ANOVA tests, and to construct the Work Ethic and Talent
scores. Excel is designed precisely for such purposes.

The online utilities used were:


• 2 Population Proportion t-Test: MedCalc proportion calculator -
https://www.medcalc.org/calc/comparison_of_proportions.php
• 2 Population Mean t-Test: QuantitativeSkills Simple Interactive Statistical Analysis -
http://www.quantitativeskills.com/sisa/statistics/t-test.htm
• 2 Population Mean t-Test: GraphPad QuickCalc -
http://www.graphpad.com/quickcalcs/ttest2/
• McCallum-Layton Confidence Interval Calculator for Means - https://www.mccallum-
layton.co.uk/tools/statistic-calculators/confidence-interval-for-mean-calculator
• Ausvet EpiTools – 2 population z-test -
http://epitools.ausvet.com.au/content.php?page=z-test-2
• Social Science Statistics –
http://www.socscistatistics.com/tests/chisquare2/Default2.aspx
• Social Science Statistics – http://www.socscistatistics.com/tests/chisquare2/Default2.aspx
• Easy Calculation – https://www.easycalculation.com/statistics/hypothesis-test-
population-mean.php

All such calculators were reputable sites, many of them specializing in statistics. The author
prefers these sites to R for these calculations due to their usability and ease of accommodating
many, many calculations.

Das könnte Ihnen auch gefallen