Statistical Analysis Report

1. Descriptive Statistics Summary

Column name count mean std min 25% 50% 75% max
Marital status 4424 1.18 0.61 1.00 1.00 1.00 1.00 6.00
Application mode 4424 18.67 17.48 1.00 1.00 17.00 39.00 57.00
Application order 4424 1.73 1.31 0.00 1.00 1.00 2.00 9.00
Course 4424 8856.6 2063.5 33.0 9085.0 9238.0 9556.0 9991.0
4 7 0 0 0 0 0
Daytime/evening 4424 0.89 0.31 0.00 1.00 1.00 1.00 1.00
attendance
Previous qualification 4424 4.58 10.22 1.00 1.00 1.00 1.00 43.00
Previous qualification 4424 132.61 13.19 95.0 125.00 133.10 140.00 190.00
(grade) 0
Nationality 4424 1.87 6.91 1.00 1.00 1.00 1.00 109.00
Mother's qualification 4424 19.56 15.60 1.00 2.00 19.00 37.00 44.00
Father's qualification 4424 22.28 15.34 1.00 3.00 19.00 37.00 44.00
Mother's occupation 4424 10.96 26.42 0.00 4.00 5.00 9.00 194.00
Father's occupation 4424 11.03 25.26 0.00 4.00 7.00 9.00 195.00
Admission grade 4424 126.98 14.48 95.0 117.90 126.10 134.80 190.00
0
Displaced 4424 0.55 0.50 0.00 0.00 1.00 1.00 1.00
Educational special 4424 0.01 0.11 0.00 0.00 0.00 0.00 1.00
needs
Debtor 4424 0.11 0.32 0.00 0.00 0.00 0.00 1.00
Tuition fees up to date 4424 0.88 0.32 0.00 1.00 1.00 1.00 1.00
Gender 4424 0.35 0.48 0.00 0.00 0.00 1.00 1.00
Scholarship holder 4424 0.25 0.43 0.00 0.00 0.00 0.00 1.00
Age at enrollment 4424 23.27 7.59 17.0 19.00 20.00 25.00 70.00
0
International 4424 0.02 0.16 0.00 0.00 0.00 0.00 1.00
Curricular units 1st 4424 0.71 2.36 0.00 0.00 0.00 0.00 20.00
sem (credited)
Curricular units 1st 4424 6.27 2.48 0.00 5.00 6.00 7.00 26.00
sem (enrolled)
Curricular units 1st 4424 8.30 4.18 0.00 6.00 8.00 10.00 45.00
sem (evaluations)
Curricular units 1st 4424 4.71 3.09 0.00 3.00 5.00 6.00 26.00
sem (approved)
Curricular units 1st 4424 10.64 4.84 0.00 11.00 12.29 13.40 18.88
sem (grade)
Curricular units 1st 4424 0.14 0.69 0.00 0.00 0.00 0.00 12.00
sem (without
evaluations)
Curricular units 2nd 4424 0.54 1.92 0.00 0.00 0.00 0.00 19.00
sem (credited)
Curricular units 2nd 4424 6.23 2.20 0.00 5.00 6.00 7.00 23.00
sem (enrolled)
Curricular units 2nd 4424 8.06 3.95 0.00 6.00 8.00 10.00 33.00
sem (evaluations)
Curricular units 2nd 4424 4.44 3.01 0.00 2.00 5.00 6.00 20.00
sem (approved)
Curricular units 2nd 4424 10.23 5.21 0.00 10.75 12.20 13.33 18.57
sem (grade)
Curricular units 2nd 4424 0.15 0.75 0.00 0.00 0.00 0.00 12.00
sem (without
evaluations)
Unemployment rate 4424 11.57 2.66 7.60 9.40 11.10 13.90 16.20
Inflation rate 4424 1.23 1.38 -0.80 0.30 1.40 2.60 3.70
GDP 4424 0.00 2.27 -4.06 -1.70 0.32 1.79 3.51

The table provides a statistical summary for various features (columns) of the student’s dataset.
1. Count: The total number of observations (rows) in the dataset for each feature. All features have
4,424 observations. This shows that there is no presence of missing values.
2. Mean: The average value of each feature.
3. Standard Deviation (std): A measure of the spread or variability in the data.
4. Min: The minimum value for each feature.
5. 25%: The 25th percentile value (first quartile), indicating that 25% of the data points are below
this value.
6. Median: The median (50th percentile), showing the middle value of the data.
7. 75%: The 75th percentile (third quartile), meaning 75% of the data points are below this value.
8. Max: The maximum value observed for each feature.
Insights:
 Marital status: Most values seem to be concentrated on a specific category, as 75% of the data
have a value of 1, while the max is 6.
 Application mode: It has a high standard deviation (17.48), with a mean of 18.67, indicating a
wide range of modes with some large outliers (max = 57).
 Application order: Most applicants appear to be in the 1st or 2nd order, as the median (50%) is 1
and the third quartile is 2. The maximum value of 9 indicates that some applicants applied in later
orders.
 Daytime/evening attendance: Most students attended during the day (median and 75th
percentile = 1, with a max of 1).
  Course: The values range widely from 33 to 9991, with an average of around 8856. The high
standard deviation (2063.57) suggests that different courses have widely varying codes or
identifiers.
 Admission grade: The average admission grade is around 127 with a standard deviation of 14.48.
The grades span from 95 to 190.
 Age at enrollment: The median age is 20 years, with a fairly large range of 17 to 70 years.
 Curricular units (1st and 2nd semesters): This feature includes the number of credited,
enrolled, evaluated, and approved curricular units, as well as the grades. The data shows a range
of involvement in curricular units across students, with most students having a median grade of
12-13 and enrolling in 6-7 units per semester.
 Unemployment rate: The unemployment rate ranges from 7.6 to 16.2, with a median of 11.1%.
 Inflation rate: The inflation rate ranges from -0.8 to 3.7, with a median of 1.4%.
 GDP: The GDP values range from -4.06 to 3.51, with a median of 0.32, indicating some
fluctuation in economic growth.

Conclusion:
 Many features, like Marital status, Daytime/evening attendance, Debtor, and Tuition fees up
to date, have mostly binary data (0 or 1).
 Some features have significant variability, such as Course, Application mode, and
Mother’s/Father’s occupation, suggesting a diverse range of values.
 Curricular units and Admission grades give insight into academic performance and
engagement, while features like Unemployment rate, Inflation rate, and GDP provide
economic context.

2. Correlation Heatmap:
Key Insights from the Heatmap:
o Curricular units 1st sem (enrolled) has a very high correlation with Curricular units
1st sem (evaluations) (0.95). This makes sense because the number of enrolled units
directly influences the number of evaluations.
o Curricular units 1st sem (evaluations) has a strong positive correlation with
Curricular units 1st sem (approved) (0.91). Students who evaluate units are likely to
pass a portion of them.
o Curricular units 2nd sem (enrolled) and Curricular units 2nd sem (evaluations)
(0.94) similarly show a very high correlation, indicating a consistent pattern of students
enrolling and being evaluated in the second semester.
o Curricular units 1st sem (approved) is strongly correlated with Curricular units 1st
sem (grade) (0.84). This is logical as better grades are generally associated with
successfully passing more units.
o Previous qualification (grade) has a moderate positive correlation with Admission
grade (0.58), suggesting that students with higher previous qualifications tend to have
better admission grades.
 o Curricular units 2nd sem (approved) correlates positively with Curricular units 2nd
sem (grade) (0.77), meaning that students who pass more courses in the second semester
tend to achieve higher grades.
o Age at enrollment has a moderate negative correlation with the Target variable (-0.25),
indicating that older students might have a lower probability of achieving the target
outcome (graduation).
o Curricular units 1st sem (approved) shows a moderately positive correlation with the
Target variable (0.54), meaning that students who pass more units in the first semester
are more likely to achieve the desired graduation rate.
o Curricular units 1st sem (grade) also correlates positively with the Target (0.49),
indicating that higher grades in the first semester contribute to better graduation rate.
o Admission grade has a positive but weaker correlation with the Target (0.12),
suggesting that students with higher admission grades are somewhat more likely to
succeed.
o Unemployment rate shows a slight positive correlation with the Target (0.08), implying
a weak relationship between unemployment and the target outcome, although the effect is
minimal.
o GDP has a very weak but positive correlation with the Target (0.04), indicating that
economic conditions may slightly influence student success.
o Curricular units 1st sem (without evaluations) has a negative correlation with the
Target (-0.085). This suggests that students who have units without evaluations in the
first semester may be less likely to meet the target.
o Curricular units 2nd sem (without evaluations) shows a stronger negative correlation
with the Target (-0.098), which implies that failing to evaluate units in the second
semester could reduce the likelihood of graduation.
o The Inflation rate has little to no correlation with other features, indicating that it
doesn’t play a significant role in student performance or any of the academic-related
features.
Conclusion:
 Academic performance in the first and second semesters, as indicated by the number of
evaluations, approvals, and grades, is highly correlated with achieving the likelihood of
graduation.
 Economic variables like GDP and Unemployment rate have minimal impact on student
outcomes.
 Students' performance in curricular units (especially in the first semester) is a strong predictor of
whether they achieve the target, while age and lack of evaluations are negatively related to
success.

3. Hypothesis Test Results:

 The implementation of the hypothesis test involves the creation of a new variable called
‘drop_stats’(dropout status) which indicates is a particular student dropped out or not. 1
represents Dropouts, 0 represents Non Dropouts irrespective of the fact that they are currently
enrolled or graduates.
For chi-square contingency table tests, the Phi’s coefficient which tell the degree and direction of
the association between the two variables was computed.

Summary of the Hypothesis Testing

Test Name Test Hypothesis Test p- Conclusi

Type Statist value on
ic
Unemployment Rate vs Correlati H0: r = 0 0.0129 0.3880 Fail to
Dropout Status on Test H1: r ≠ 0 8 79 reject H0
Gross Domestic Product vs Correlati H0: r = 0 0.0462 0.0020 Reject
Dropout Status on Test H1: r ≠ 0 3 59 H0
Inflation Rate vs Dropout Correlati H0: r = 0 0.0642 0.0642 Fail to
Status on Test H1: r ≠ 0 2 23 reject H0
Test of Association Chi- H0: The variables are 232.82 3.25E- Reject
between Debt status and Square independent 5 49 H0
Dropout status Test H1: The variables are not
independent
Test of Association Chi- H0: The variables are 814.76 4.9E- Reject
Between Tuition fees status Square independent 4 175 H0
and Dropout status Test H1: The variables are not
independent
Admission grade for T-Test H0: μ1 = μ2 -6.237 2.59E- Reject
dropouts vs non dropouts H1: μ1 < μ2 10 H0
Test of Association Chi- H0: The variables are 266.31 1.98E- Reject
between Scholarship Square independent 8 56 H0
holder and Dropout status Test H1: The variables are not
independent

i. Unemployment Rate vs Dropout Status:

There is no evidence to suggest a significant relationship between the unemployment
rate and the likelihood of dropping out. Therefore, we fail to reject the null
hypothesis that the correlation is zero, implying that unemployment rate does not
seem to influence student dropout rates based on this data.
ii. GDP vs Dropout Status:
There is evidence to suggest a weak but statistically significant correlation between
GDP and dropout rates. Based on the data, we reject the null hypothesis (H0),
implying that GDP does have a small effect on the likelihood of students dropping
out.
iii. Inflation Rate vs Dropout Status:
There is evidence to suggest a weak but statistically significant correlation between
GDP and dropout rates. Based on the data, we reject the null hypothesis (H0),
 implying that GDP does have a small effect on the likelihood of students dropping
out.
iv. Debt Status vs Dropout Status:
This means that there is a strong association between debt status and dropout status.
Phi's coefficient measures the strength of the association between the two variables. A
value of 0.22941 indicates a moderate positive association. This means that as debt
status increases, the likelihood of dropping out also tends to increase.
v. Tuition fees payment status vs Dropout Rate:
Given the p-value is much smaller than 0.05, we reject the null hypothesis (H0). This
implies that there is a statistically significant relationship between "Tuition fees up to
date" status and "Dropout status" — the two variables are not independent.
The Phi's Coefficient of -0.42915 indicates a moderate negative association between
the two variables. The negative sign suggests that individuals with up-to-date tuition
fees are less likely to drop out, while those with unpaid fees are more likely to drop
out.
vi. Admission grade of Dropouts vs Admission grade of Non Dropouts:
The p-value (2.59e-10) is extremely small and well below the significance level (α =
0.05), indicating that the difference is statistically significant.
The evidence strongly supports the alternative hypothesis. We reject the null
hypothesis (H0) and conclude that students who dropped out had significantly lower
admission grades compared to those who did not.
vii. Scholarship Holder vs Dropout Status:
The extremely small p-value (1.983E-56 is far below the significance level (α =
0.05), indicating a highly significant result.
Phi's coefficient is a measure of association strength for Chi-square tests. A value of -
0.24535 suggests a moderate negative association between the two variables (e.g., as
scholarship rate increases, the likelihood of dropout may decrease).