You are on page 1of 17

Advances and Applications in Statistics

© 2023 Pushpa Publishing House, Prayagraj, India


http://www.pphmj.com
http://dx.doi.org/10.17654/0972361723013
Volume 86, Number 1, 2023, Pages 1-17 P-ISSN: 0972-3617

THE ABILITY OF ADMISSION CRITERIA IN


PREDICTING STUDENT’S SUCCESS IN THE
ENGINEERING PATH AT THE UNIVERSITY
OF HAFR AL-BATIN

Abstract

This article analyzes how the admission criteria of the University of


Hafr Al-Batin can predict the first-year college students’ performance
in the engineering path through their first-year college GPA (FYGPA),
by utilizing the high school GPA (HSGPA) and the standardized test
scores in the Kingdom of Saudi Arabia (ACT and APT). To compare
the models, the Bayesian information criterion (BIC) and the Akaike
information criterion (AIC) are used as model selection criteria. The
validity of the entrance tests is supported by actual data.

Received: December 12, 2022; Accepted: January 25, 2023


2020 Mathematics Subject Classification: 62J05, 62J20, 62H20.
Keywords and phrases: regression, Bayesian information criterion, Akaike information
criterion, GPA, college admission.
How to cite this article: Lulah Alnaji, The ability of admission criteria in predicting student’s
success in the engineering path at the University of Hafr Al-Batin, Advances and Applications
in Statistics 86(1) (2023), 1-17. http://dx.doi.org/10.17654/0972361723013
This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Published Online: February 4, 2023
2 Lulah Alnaji
1. Introduction

There has been a move in recent decades toward making standardized


testing a prerequisite for university entrance. There are several studies
conducted all over the world to determine the efficacy of admission tests
in forecasting students’ likelihood of success and predicting their college
GPA based on admission exams - high school GPA (HSGPA), ACT score,
and APT score. Examining student populations across many institutions
has been the focus of a certain number of studies. Finding an overall mean
correlation between the admission test scores and HSGPA is challenging
when analyzing the validity of the entrance test scores for numerous
institutes and the data need to be combined in an observant way. For
instance, [1] identified the correlations of each institute independently,
corrected the correlations with correction method in [2], and then estimated
the mean correlation across several institutes. According to [1]’s report in
making a comparison of private and public organizations, they displayed the
findings of a thorough investigation of the reliability of admissions exams
and found that the correlation between the admission tests in conjunction
with HSGPA and (FYGPA) is a little bit higher for private schools.

Multiple correlations of a number of institutions, and the various


correlations between entrance test results and FYGPA of student populations
across a number of institutes have been investigated in [3]. According to the
findings of [3], the correlation between entrance test results and FYGPA of
student populations in institutions with a wider variety of extracurricular
activities are lower for those residing in the institutes’ housing. The report
has a number of elements that were focused on by [3]. For instance, the
study showed a greater correlation between the admission test scores and
FYGPA for liberal arts colleges when compared to other colleges included
in the report. Moreover, in terms of comparing small towns and big cities,
small towns’ institutes have a higher correlation.

Cohn et al. [4] investigated student populations over more than one
campus of the University of California, and came to the conclusion that the
The Ability of Admission Criteria in Predicting Student’s Success … 3
correlation between admission test scores and FYGPA varies from one
campus to another. Although studies covering a number of universities are
useful, they do not elaborately specify information on the extent to which
admission tests relate to student performance over a number of institutes,
and do not specify the associated influences to the variability.

It can be challenging to generalize the findings of research on the


reliability of standardized exams and HSGPA in predicting students’ success
across a range of institutions. Each institute has multiple layers of student
population clustering, including HSGPA, program, college, and many others.

Additionally, the procedures for admission to different colleges within


the same university have different standards and test procedures.

Numerous institutes carried out regional studies to provide more in-


depth details on the best predictor of college students’ GPAs. Work has been
done for particular courses to look at the correlation between specific
courses’ GPAs and the admission test scores. The correlation between a
particular department and admission test scores has also been studied. The
best predictor of admission criteria (the average achievement test score
(ACH), high school class rank, and the SAT), and also the best predictor of
student achievement at the University of Pennsylvania have been presented
in [5]. The findings show that high school class rank and ACH were more
accurate predictors of students’ performance than the SAT.

The GPA of more than 5,000 students at the University of California,


San Diego has been studied in [6]. The paper concluded that the HSGPA and
SAT both add significantly to the prediction of the students’ performance.

The extent of the ability of the HSGPA and SAT in predicting success
at the University of South Carolina, using a sample of 521 undergraduates
enrolled in some classes of Principles of Economics courses had been a
matter of investigation in [7].

Before the Saudi Ministry of Education started moving toward


using standardized examinations as entrance requirements for universities,
4 Lulah Alnaji
admittance to Saudi universities was dependent on the HSGPA [8].
Standardized tests like the ACT and APT as well as the high school grade
point average are used by Saudi universities to determine entrance. Given
that all students are evaluated using the same materials and under the same
settings, the ACT and APT examinations are frequently considered by Saudi
universities to be stronger and more accurate predictors of success than the
HSGPA [9]. The ACT test is a comprehensive examination of high school
courses in biology, chemistry, physics, mathematics, and english, while
the APT test’s result is based on the verbal and quantitative components.
For further information, see [9].

Due to the large number of graduates from secondary schools and the
limited number of seats in Colleges of Engineering (College of Engineering
and College of Computer Engineering), the University of Hafr Al-Batin
resorts to placing a competition among students for seats based on the grades
of standardized tests and HSGPA, to be admitted first to the engineering path
for a year and based on the FYGPA, the University of Hafr Al-Batin makes a
decision. Students with high GPAs can be enrolled in a department of one of
the two colleges according to the student’s own choice, and the others are
given an opportunity to raise the GPAs or transfer to another college.

To the best of our knowledge, no investigation of the effectiveness of the


requirements for admission to engineering colleges at the University of Hafr
Al-Batin has been done or has been made publicly available. As a result, this
study is the first to look into the effectiveness of standardized tests and
HSGPA in forecasting FYGPA for students in the Colleges of Engineering at
the University of Hafr Al-Batin.

The remainder of this article is divided into the following sections. The
descriptive statistics and correlations for the actual data of both the genders
at the engineering colleges are shown in Section 2. Section 3 addresses the
validity of the ACT, APT, and HSGPA in predicting FYGPA and provides
the findings of the inferential statistics. The conclusion is provided in
Section 4.
The Ability of Admission Criteria in Predicting Student’s Success … 5
2. Descriptive Statistics and Correlations

The University of Hafr Al-Batin, a public university, served as the site of


this study. The university’s decision support and performance control unit
provided the study’s data (HSGPA, APT score, ACT score, FYGPA). 473
male and 426 female students who were admitted to the four-year college in
the fall of 2019 made up the sample. The GPA was calculated using a 4.00
scale at the University of Hafr Al-Batin. The analysis of the data was done
using the RStudio software.

Figure 1. The plot of percentage of the sample data.

2.1. Descriptive statistics

Descriptive statistics for this study sample are summarized in Tables 1


to 3 for female and male students, respectively, to obtain the descriptive
information (minimum, maximum, 1. quartile, 3. quartile, median, mean,
std. dev., skewness, and kurtosis) about the data used in this study. Tables 2
and 3 display that the minimum for the ACT, HSGPA, and FYGPA are
higher for female students, while the APT is slightly higher for male
students. Also, the maxima of the standardized tests (ACT and APT) are
6 Lulah Alnaji
higher for female students. In general, descriptive information such as
(minimum, maximum, 1. quartile, 3. quartile, median, mean and std. dev.)
which are displayed in Tables 2 and 3 are higher for female students in most
cases than male students.

For the skewness, Tables 1 to 3 and Figures 2 to 6 demonstrate that the


standardized tests (ACT and APT) and FYGPA for female and male students
are distributed normally, while the HSGPA is negatively skewed for both the
genders, implying that there are some HSGPAs significantly far away from
most of the other HSGPAs. It is clear from Table 1 that the mean of HSGPA
for all students is (92) and the median is (94) while the minimum of the
HSGPA is (70), and the maximum is closer to the mean and the median
which showed as a longer tail on the left side of the distribution of the
data of this study. For female students, Table 2 displays the mean, median,
minimum and maximum HSGPA as (95), (96), (84) and (100), respectively.
Finally, HSGPA for male students as demonstrated in Table 3, the mean
and the median are (90) and (91), respectively, while the minimum and the
maximum are (70) and (100), respectively, which provide a non-symmetrical
graph.

The means for the FYGPA for both the genders are typically lower than
the means for HSGPA. Table 2 exhibits that the mean of HSGPA is 95
which considers (A) as a grade, corresponding to grade (B) as the mean of
FYGPA. Also, Table 3 shows that the mean for the male students of
HSGPA is 90.57 (A) while the mean for FYGPA is 2.9 (i.e., (B)), signifying
that the engineering path studies at the University of Hafr Al-Batin are
tougher than high-schools studies.
The Ability of Admission Criteria in Predicting Student’s Success … 7
Table 1. Descriptive statistics of the data for all students
Descriptive statistics ACT APT HSGPA FYGPA
No. obs. 899.000000 899.000000 899.000000 899.000000
Minimum 53.000000 59.000000 70.000000 1.600000
Maximum 98.000000 100.000000 100.000000 4.000000
1. Quartile 67.000000 70.000000 90.000000 2.695000
3. Quartile 81.000000 81.000000 97.000000 3.480000
Mean 74.218020 75.438265 92.802002 3.050044
Median 73.000000 75.000000 94.000000 3.040000
SE mean 0.296361 0.244162 0.177485 0.017468
LCL mean 73.636379 74.959070 92.453669 3.015762
UCL mean 74.799661 75.917459 93.150336 3.084327
Variance 78.959096 53.593902 28.319328 0.274301
Std. dev. 8.885893 7.320786 5.321591 0.523737
Skewness 0.227343 0.183677 -1.023225 -0.262314
Kurtosis -0.635306 -0.663955 1.008724 -0.496278

Table 2. Descriptive statistics of the data for female students


Descriptive statistics ACT APT HSGPA FYGPA
No. obs. 427.000000 427.000000 427.000000 427.000000
Minimum 57.000000 59.000000 84.000000 1.680000
Maximum 98.000000 100.000000 100.000000 4.000000
1. Quartile 71.000000 71.000000 93.000000 2.810000
3. Quartile 85.000000 83.000000 98.000000 3.575000
Mean 77.906323 76.672131 95.257611 3.165995
Median 78.000000 76.000000 96.000000 3.190000
SE mean 0.410030 0.373370 0.158394 0.024873
LCL mean 77.100389 75.938254 94.946281 3.117106
UCL mean 78.712258 77.406008 95.568942 3.214885
Variance 71.789326 59.526052 10.712823 0.264174
Std. dev. 8.472858 7.715313 3.273045 0.513978
Skewness 0.042409 0.084085 -0.631056 -0.389549
Kurtosis -0.982984 -0.765242 0.044392 -0.504525
8 Lulah Alnaji
Table 3. Descriptive statistics of the data for male students
Descriptive statistics ACT APT HSGPA FYGPA
No. obs. 472.000000 472.000000 472.000000 472.000000
Minimum 53.000000 60.000000 70.000000 1.600000
Maximum 95.000000 90.000000 100.000000 4.000000
1. Quartile 65.000000 69.000000 87.000000 2.600000
3. Quartile 76.000000 79.000000 95.000000 3.300000
Mean 70.851380 74.305732 90.573248 2.942972
Median 70.000000 74.000000 91.000000 2.940000
SE mean 0.362314 0.311442 0.268434 0.023456
LCL mean 70.139424 73.693742 90.045769 2.896880
UCL mean 71.563336 74.917723 91.100728 2.989065
Variance 61.828929 45.685052 33.938772 0.259143
Std. dev. 7.863137 6.759072 5.825699 0.509061
Skewness 0.361542 0.185538 -0.614839 -0.196070
Kurtosis -0.144208 -0.669165 0.051927 -0.409486

Figure 2. The plot of standardized tests for all students.


The Ability of Admission Criteria in Predicting Student’s Success … 9

Figure 3. The plot of standardized test (ACT) for all students by gender.

Figure 4. The plot of standardized test (APT) for all students by gender.
10 Lulah Alnaji

Figure 5. The plot of standardized tests (ACT and APT) for all students by
gender.

Figure 6. The plot of GPAs (HSGPA and FYGPA) for all students by
gender.

2.2. Correlations

This study’s primary goal is to demonstrate the correlation between the


standardized test in conjunction with HSGPA and the FYGPA for students
in the engineering path. Additionally, gender-specific (Tables 5 and 6)
correlation studies are undertaken after the general correlation analysis
(Table 4). For the overall correlation analysis, in Table 4, we find that
ACT has a stronger correlation r  0.62 with the FYGPA, followed by the
other standardized test APT r  0.54 , while the HSGPA has the weakest
correlation r  0.40 among all.
The Ability of Admission Criteria in Predicting Student’s Success … 11
For gender-specific correlation analysis, Table 5 shows a little variation
of the female students’ correlation results of ACT r  0.67 and APT
r  0.65 , in their relationship with FYGPA. However, there is a stronger
correlation between ACT and FYGPA than APT and HSGPA. For male
students, things changed a bit as shown in Table 6. The correlation analysis
still demonstrates that ACT has the strongest correlation r  0.52 with
HSGPA. There is a slight difference in the male students’ correlation
analysis of APT r  0.32 and HSGPA r  0.39 , but, on the other hand,
ACT and FYGPA have a stronger correlation than APT and HSGPA, in
contrast to the tables above (Tables 4 and 5).

Table 4. Correlation between the inputs (ACT, APT, HSGPA) and the
outcome (FYGPA) for male and female students
Correlation ACT APT HSGPA
FYGPA 0.6170954
FYGPA 0.5400429
FYGPA 0.3934539

Table 5. Correlation between the inputs (ACT, APT, HSGPA) and the
outcome (FYGPA) for female students
Correlation ACT APT HSGPA
FYGPA 0.6737089
FYGPA 0.6487951
FYGPA 0.4329735

Table 6. Correlation between the inputs (ACT, APT, HSGPA) and the
outcome (FYGPA) for male students
Correlation ACT APT HSGPA
FYGPA 0.518568
FYGPA 0.3167674
FYGPA 0.3959756
12 Lulah Alnaji

Figure 7. The plots of the relationships between FYGPA and ACT, APT,
HSGPA, respectively, for male students.

Figure 8. The plots of the relationships between FYGPA and ACT, APT,
HSGPA, respectively, for male students.
The Ability of Admission Criteria in Predicting Student’s Success … 13

Figure 9. The plots of the relationships between FYGPA and ACT, APT,
HSGPA, respectively, for male students.

3. Inferential Statistics

The correlation analysis in the previous Subsection 2.2 provides an


important message that ACT is a better predictor than APT and HSGPA.
That needs to be confirmed by regression analysis as well, which is provided
in this section. Table 7 provides the regression analysis for the data of this
study, while Table 8 shows the analysis by gender. The effects of the
predictors (ACT, APT, HSGPA) on the measures of first-year college
performance are estimated in Table 7 by using the models below:

Y    1 ACT  , (1)

Y  0  1 APT  , (2)

Y  0  1HSGPA  , (3)

Y  0  1 ACT  2 APT  , (4)

Y  0  1 ACT  2 APT  3HSGPA  , (5)


14 Lulah Alnaji
where Y and  are the dependent variable (FYGPA) and the random error,
respectively.

The intercepts represent the average FYGPA when the total percentile
score is 0 for the predictors. Obviously, no student has a zero score, so it
does not make sense to interpret the intercepts in these particular regression
models (1)-(5). Consequently, they can be omitted safely. Regarding single-
variable models (1)-(3), the slopes indicate that, on average, the FYGPA
tends to rise by 0.036, 0.039 and 0.039 for every rise in the ACT, APT,
HSGPA of one, respectively. It is clear that the slope values of the models
(1)-(3) have slight to no difference, but using R 2 to compare the models
(1)-(3), it can be seen that in the model (1) ACT accounts for 38% of the
variation in FYGPA, whereas model (2) accounts for 29% of it, and finally,
model (3) illustrates only 15% of the variation of FYGPA.

For model (4), 1  0.027 of ACT suggests that the model forecasts
an average improvement in the HSGPA of about 0.027 points if the ACT
score rises by one percentile, controlling for APT scores. In contrast, an
improvement of one percentile in the APT score corresponds to an average
rise of 2  0.019 points in the FYGPA, controlling for ACT scores. For
model (5) in Table 7, 1  0.023 of ACT states that the model forecasts an
average gain in the HSGPA of about 0.023 per point for every percentile
improvement in the ACT score, controlling for APT scores and HSGPA.
In contrast, adjusting for ACT scores and HSGPA, an increase in APT score
of one percentile corresponds to an average 0.019 point rise in FYGPA,
whereas an increase in HSGPA of one percentile corresponds to an average
0.017 point change in FYGPA. Using R 2 Adj. to compare the models (4)-(5),
it can be seen that in the model (4) ACT accounts for 43% of the variation in
FYGPA, whereas model (5) accounts for 45% of it. Moreover, making the
compression between models (1) and (5) using AIC and BIC criteria reveals
that model (5), which includes all predictor variables, has the lowest AIC
and BIC, leading to result that it is the best model among the others.
The Ability of Admission Criteria in Predicting Student’s Success … 15
Table 7. Regression analysis for all students
Model coeff. Model (1) Model (2) Model (3) Model (4) Model (5)
No. obs. 899 899 899 899 899
(Intercept) 0.351 0.135 -0.544 -0.387 -1.689
ACT 0.036 - - 0.027 0.023
APT - 0.039 - 0.019 0.019
HSGPA - - 0.039 - 0.017
R2 0.381 0.292 0.155 0.426 0.453
2
R Adj. 0.380 0.291 0.154 0.425 0.451
AIC 962.4 1083.4 1242.2 896.3 855.3
BIC 976.8 1097.8 1256.6 915.5 879.4

Using the following models, to observe the variations in FYGPA are


influenced by taking gender into account:

Y  0  1 ACT  2Gender  , (6)

Y  0  1 APT   2Gender  , (7)

Y  0  1HSGPA  2Gender  , (8)

Y  0  1 ACT  2 APT  3Gender  , (9)

Y  0  1 ACT  2 APT  3HSGPA   4Gender  , (10)

where Y and  are the dependent variable (FYGPA) and the random error,
respectively.

When the entire percentile score for the predictors is zero, the intercepts
indicate the average FYGPA. It is clear that no student has a score of zero,
hence, it is pointless to interpret the intercepts in these specific regression
models (6)-(10). As a result, it is safe to omit them.

When gender is taken into consideration, a one percentile shift in the


ACT score corresponds, on average, to a 0.04 point change in FYGPA in
each of models (6)-(8). For male students, the average increases in their
FYGPA after adjusting for ACT scores by 0.048 points. In contrast, males’
16 Lulah Alnaji
FYGPAs fall, on average, by 0.13 points when APT scores are controlled,
also, when HSGPA results are taken into account, males’ FYGPAs fall,
on average, by 0.048 points. R 2 Adj. -based model comparisons reveal that
model (10) explains most of the variations in FYGPA than other models
(6)-(10). Moreover, (AIC and BIC)-based model comparisons confirm that
(10) is the best model since it has the lower values than models (6)-(9).

Table 8. Regression analysis for all students by gender


Model coeff. Model (6) Model (7) Model (8) Model (9) Model (10)
No. obs. 899 899 899 899 899
(Intercept) 0.244 0.315 -0.334 -0.431 -2.120
ACT 0.037 - - 0.028 0.025
APT - 0.037 - 0.019 0.018
HSGPA - - 0.037 - 0.021
Gender M 0.048 -0.130 -0.048 0.024 0.101
R2 0.383 0.307 0.156 0.426 0.460
R2 Adj. 0.381 0.305 0.155 0.425 0.457
AIC 961.9 1066.3 1242.4 897.6 846.2
BIC 981.1 1085.5 1261.6 921.6 875.0

4. Conclusion

Considering that HSGPA and APT have varying impacts on FYGPA


depending on a student’s gender, the ACT is a stronger predictor of first-year
GPA than APT and HSGPA for the FYGPA for engineering students. This
article explores the academic elements that are generally employed in
college entrance applications and are used by a university to determine
admission in order to have a better knowledge of the relationships
between academic ability and academic performance. Nonacademic
elements including family income, parents’ educational backgrounds, and
environmental factors might be included in future research. Additionally, we
might investigate the correlation of the graduate GPA in comparison to the
FYGPA.
The Ability of Admission Criteria in Predicting Student’s Success … 17
References

[1] Leonard L. Baird, Predicting predictability: the influence of student and


institutional characteristics on the prediction of grades, ETS Research Report
Series 1983(2) (1983), i-11.

[2] Julian R. Betts and Darlene Morell, The determinants of undergraduate grade
point average: the relative importance of family background, high school
resources, and peer group effects, Journal of Human Resources 34 (1999),
268-293.

[3] Jonathan Baron and M. Frank Norman, SATs, achievement tests, and high-school
class rank as predictors of college performance, Educational and Psychological
Measurement 52(4) (1992), 1047-1055.

[4] Elchanan Cohn, Sharon Cohn, Donald C. Balch and James Bradley, Jr.,
Determinants of undergraduate GPAs: SAT scores, high-school GPA and high-
school rank, Economics of Education Review 23(6) (2004), 577-586.

[5] Establishment of National Center for Assessment in Higher Education. At:


https://etec.gov.sa/en/productsandservices/Qiyas/Education/Pages/default.aspx.

[6] Jennifer L. Kobrin, Brian F. Patterson, Emily J. Shaw, Krista D. Mattern and
Sandra M. Barbuti, Validity of the SAT® for predicting first-year college grade
point average, Research Report No. 2008-5, College Board, 2008.

[7] D. N. Lawley, IV.—A note on Karl Pearson’s selection formulae, Proceedings of


the Royal Society of Edinburgh Section A: Mathematics 62(1) (1944), 28-30.

[8] National Center for Assessment in Higher Education. At:


https://etec.gov.sa/en/About/Centers/Pages/qiyas.aspx.

[9] Rebecca Zwick, Terran Brown and Jeffrey C. Sklar, California and the SAT: A
Reanalysis of University of California Admissions Data, Research and Occasional
Paper Series: CSHE, 8.04, Center for Studies in Higher Education, 2004.

You might also like