New chatGPT 3.5 Instruction (Prompt) to Calculate Statistical

Indicators for Student Graduation Projects

VALERY OKULICH-KAZARIN

National Louis University,

Nowy Sącz,

POLAND

Abstract: - The paper aims to develop a new chatGPT 3.5 instruction (prompt) for computing statistical

indicators in student graduation projects. A bibliometric analysis of 79606 sources published in the Scopus

database revealed a high level of interest in solving problems related to "graduation projects" and "statistical

indicators." Numerous studies emphasize the importance of probability and statistics education. Concurrently,

educators are advised to abandon teaching manual calculation methods to students. ChatGPT could serve as a

modern tool for computing statistical indicators. Modern methods employed in this research included reviewing

scientific literature, analysis and synthesis, bibliometric analysis, mathematical modeling, computation of

statistical indicators, and verification of statistical hypotheses using Z-statistics. Five examples of calculating

statistical indicators are provided in this paper. Three tools were used for computing statistical indicators, with

the new chatGPT 3.5 instruction (prompt) serving as the experimental method, while Excel tables and

Windows calculator were used as control methods. Verification of statistical hypotheses using Z-statistics

demonstrated the equality of results between experimental and control methods. The standard testing level was

set at α = 0.05. The novelty of this work lies in the creation of the new chatGPT 3.5 instruction (prompt) for

computing statistical indicators in student graduation projects. Additionally, a User's Guide has been published.

The practical value of this work lies in reducing the time and simplifying the method for computing statistical

indicators in preparing graduation projects, as well as in improving their quality. An additional benefit is the

expanded use of computers for educational purposes.

Key-Words: - chatGPT 3.5, prompt, computer, graduation project, statistical indicators, student.

5HFHLYHG$SULO5HYLVHG)HEUXDU\$FFHSWHG0DUFK3XEOLVKHG0D\

1 Introduction

The development of information technologies leads

to growth the possibility of implementing new tools

in education, [1], [2], [3], [4], [5], [6], [7],

mathematical modeling, [8], [9] and the use of

intelligent systems, [10], [11], [12]. For instance,

ChatGPT provides users with the ability to interact

with a computer using natural language. This makes

ChatGPT a valuable tool in educational processes.

Currently, such tools as ChatGPT are becoming

increasingly popular in various fields, including

higher education, [12], [13], [14], [15], [16].

"ChatGPT is an AI-powered language model

developed by OpenAI. It has been trained on a

massive amount of text data from the internet and

can generate human-like text responses to a given

prompt. It can answer questions, converse on a

variety of topics, and generate creative writing

pieces" (https://chatgpt.org/). Model 3.5 is a

publicly available free model. Therefore, it was

chosen in this study.

This article is dedicated to a new guide on using

ChatGPT version 3.5 for computing statistical

indicators in student graduation projects.

In the modern educational process, student

graduation projects play a significant role in the

development and assessment of students' skills.

Computing statistical indicators is a key stage in the

analysis of empirical data. Therefore, scientists from

different countries are researching the process of

teaching students the methodology of computing

statistical indicators, [17], [18], [19], [20], [21],

[22], [23], [24], [25], [26].

This article presents an instruction for

computing statistical indicators, such as the mean

and standard deviation, using the neural network

ChatGPT version 3.5. The possibilities of using

various software tools and technologies, including

artificial intelligence, for automating the process of

computing and presenting statistical data are

discussed.

The use of ChatGPT 3.5 in student graduation

projects opens up new perspectives for students.

This research enables students to more effectively

WSEAS TRANSACTIONS on COMPUTER RESEARCH

DOI: 10.37394/232018.2024.12.30

Valery Okulich-Kazarin

E-ISSN: 2415-1521

307

Volume 12, 2024

apply statistical methods in their graduation

projects, thus contributing to the development of

their professional skills and the improvement of the

quality of their final work.

The goal was to provide students with an

instruction (prompt) for effectively using this

technology to compute statistical indicators in their

graduation projects. This instruction (prompt) can

also be used in other research and practical projects

of students. The author will present the prompt and

a guide for its application. A very important

advantage of the new chatGPT 3.5 prompt is the

rejection of entering an array of data for computing

statistical indicators.

Additionally, the author will provide 5 examples

of its usage to simplify the processing of statistical

data in student projects.

The author hopes that this instruction (prompt)

will become a valuable resource for students,

helping them reduce the time spent on preparing

their graduation projects and improve their quality.

2 Problem Formulation

Works, [27], [28], [29], demonstrate that graduation

projects as a didactic tool have been endorsed by

practitioners and researchers in the field of higher

education. They value this didactic tool for its rigor,

depth of content, and engagement of students in

cutting-edge academic work. Student graduation

projects reflect a causal model of behavior, the

ability to exchange knowledge, and teamwork. Such

projects have had a positive impact on graduates.

The results of the first step of a bibliometric

analysis are visualized in Figure 1. The analysis was

conducted using 2590 document search results from

the last 30 years (from 1994 to 2023). Document

search was performed using the keywords

"graduation project" in the Scopus database

(https://www.scopus.com/term/analyzer.uri).

Fig. 1: Number of documents from 1994 to 2023 by

topic «graduation project»

Figure 1 illustrates a generally unstable growth

trend in the number of publications on the topic of

"graduation project" from the beginning to the end

of the analyzed period. The overall trend of

increasing publication numbers indicates a growing

interest among researchers in the topic of

"graduation project." Figure 1 provides indirect

evidence of existing issues in the field of

"graduation project" that require increasing attention

from researchers.

The second stage of the bibliometric analysis is

visualized in Figure 2. Here, 72774 document

search results from the last 30 years (from 1994 to

2023) were processed. Document search was

conducted using the keywords "statistical

indicators" in the Scopus database

(https://www.scopus.com/term/analyzer.uri).

Fig. 2: Number of documents from 1994 to 2023 by

topic «statistical indicators»

Figure 2 depicts a consistent and stable growth

trend in the number of publications on the topic of

"statistical indicators" from the beginning to the end

of the analyzed period. The increasing trend in

publication numbers indicates a growing interest

among researchers in the topic of "statistical

indicators." This suggests that researchers are

continuously working on addressing issues in the

field of "statistical indicators" (Figure 2).

Numerous scientific works emphasize the

importance of this topic. For example, the

importance of studying probability and statistics is

recognized in the current Australian mathematics

curriculum, [21]. Meanwhile, work [23], asserts that

psychology students often experience anxiety about

studying statistics, which can affect their

performance. Simultaneously, the author of paper

[25], states that sociology students may reluctantly

engage in the study of statistical methods with

apprehension. Furthermore, statistics courses are

perceived by students as challenging, [17]. The

attitude, motivation, and preparedness of students

can negatively impact their experience and

WSEAS TRANSACTIONS on COMPUTER RESEARCH

DOI: 10.37394/232018.2024.12.30

Valery Okulich-Kazarin

E-ISSN: 2415-1521

308

Volume 12, 2024

potentially hinder students' learning of statistics,

[17].

One of the reasons for students' negative attitude

towards studying statistics is the limited amount of

time dedicated to diversity in courses. Additionally,

teachers refer to barriers such as lack of relevance

and insufficient time, [26]. Integrating content into

statistics is a challenging task, however, the efforts

are worthwhile. Therefore, integrating diverse

content into statistics courses is highly beneficial,

[26].

Paper [19], notes that statistics textbooks

typically teach manual methods of computing

statistical indicators. The authors recommend that

teachers move away from teaching manual

computations and focus on conceptual information

and software applications, [19].

Thus, the research problem was justified by the

increasing interest in solving issues in the fields of

"graduation project" and "statistical indicators." The

results of previous research directly guide

researchers in seeking a method of computing

statistical indicators without manual computations.

This implies the need to create new information

tools for such computations.

On the third stage of the bibliometric analysis, the

author visualized 4242 document search results for

the keyword "chatGPT" (Figure 3). Document

search was conducted in the Scopus database

(https://www.scopus.com/term/analyzer.uri) over

the last 30 years (from 1994 to 2023).

Unfortunately, the Scopus database only provided

results for the period 2015-2023.

Fig. 3: Number of documents from 2015 to 2023 by

topic «chatGPT»

Figure 3 illustrates a consistent absence of

publications on this topic until 2022. However, in

2023, there were 4233 scientific documents related

to chatGPT published. This indicates a sharp

increase in research interest in this topic (Figure 3).

Summarizing the results of the bibliometric

analysis (Figure 1, Figure 2 and Figure 3), you will

notice three new scientific facts:

1. The interest of researchers in the topic of

“graduation project” can be described by an

unsustainably growing trend line.

2. The interest of researchers in the topic of

“statistical indicators” can be described by a

steadily growing trend line.

3. Research interest in the topic “chatGPT” is

characterized by a sharp increase from zero in

2022 to more than 4233 publications in 2023.

As a result, a joint analysis of these three

Figures may indicate a high scientific novelty of the

idea of using chatGPT to perform statistical

calculations when completing graduate work. Figure

1 and Figure 2 prove the high practical significance

of this article.

The author chose to compute statistical

indicators using chatGPT 3.5. When a student uses

the new chatGPT prompt for computing statistical

indicators, they save time on manual computations.

This time can be allocated to focusing on conceptual

information and software applications, as

recommended, [19].

Why did the author choose version 3.5? This

version of chatGPT was chosen due to its free

availability and accessibility for students. The

possibility of errors in calculations using chatGPT

3.5 was also confirmed in the study, [30].

The research goal is to create a new chatGPT

3.5 instruction (prompt) to calculate statistical

indicators for student graduation projects.

The research goal was achieved using modern

methods: review of scientific sources, analysis and

synthesis, bibliometric analysis, mathematical

modeling, calculation of statistical indicators,

verification of statistical hypotheses (Z-statistics).

3 Problem Solution

The research was performed from December 2023

to February 2024. The empirical part of the study

was carried out at National Louis University. For

calculations, standard tools were used that are used

in the normal educational process (including the

implementation of graduation projects) of the named

university. The author used standard package for

Microsoft Office Excel 2007 and Windows 10.

After the problem formulation, the problem was

solved in five steps. The steps outlined in the article

are as follows:

Step 1: Search for important sample sizes by

mathematical modeling.

Step 2: Creation of the new chatGPT 3.5

instruction (prompt) for computing statistical

indicators.

WSEAS TRANSACTIONS on COMPUTER RESEARCH

DOI: 10.37394/232018.2024.12.30

Valery Okulich-Kazarin

E-ISSN: 2415-1521

309

Volume 12, 2024

Step 3: Check of the new chatGPT 3.5 instruction

(prompt) using 5 examples, including 2 real cases

and 3 abstract cases.

Step 4: Execution of Z-statistics to compare the

results of calculations of statistical indicators

obtained by the new prompt and by using Windows.

Step 5: Writing a user's guide to the new chatGPT

3.5 instructions (prompt) for students conducting

graduation projects.

It's noted that the numbering of the steps in the

text has been changed.

3.1 Sample Sizes

The first step involved calculating 5 sample size

values for which modeling the instruction (prompt)

to calculate statistical indicators for student

graduation projects would be performed.

The logic of this step is based on the experience

of supervising student graduation projects (works).

The maximum size of the population under

investigation by students when conducting

graduation projects was taken as the basis. In

practice, this population size was assumed to be less

than 50000 units. The minimum number of units for

a master's graduation project should be around 200.

So, these sample sizes (200-50000) are considered

sufficient for a graduation project.

It is considered that for making business

decisions, the sampling error should not exceed

4.00%, [31]. The standard testing levels of 0.95 or

0.99 are most commonly used, [31]. In our case, a

sampling error of 4.00% and the standard testing

level of 0.95 were chosen.

Table 1 shows the sample size values for further

calculations. Mathematical modeling is performed

using an online calculator, [32].

Table 1. Results of mathematical modeling for

choosing the sample sizes

General

population size

200

500

1000

3000

50000

Sample size, N

150

273

375

500

593

Table 1 shows that with a population size of

50,000 units, the sample size is N=593 units.

Increasing the population size to 100000 units

resulted in a sample size of N=597 units, [32]. For

populations exceeding 100000 units, the sample size

is N=600 units, [32]. This means that the chosen

maximum sample size of N=593 units is justified for

graduation projects.

Additionally, Table 1 displays 5 sample size

values for which calculations of statistical indicators

M(x) and δx will be performed, [33], [34]. In other

words, further calculations will be conducted for

five sample size values (Table 1).

3.2 Creating the New chatGPT 3.5

Instruction (Prompt) to Calculate

Statistical Indicators

As mentioned earlier, chatGPT 3.5 easily performed

computing statistical indicators for sample sizes of

less than 10 units. However, for sample sizes

exceeding 10 units, chatGPT 3.5 occasionally

produced errors in the results. As shown in Table 1,

the sample size theoretically can reach up to 593

units in graduation projects.

When creating the instruction (prompt), the

following simple scenarios were considered:

students conduct a basic study where only 2 states

need to be evaluated (agree - disagree; on - off; yes -

no; black - white; growth-decrease; etc.). These are

situations where responses can be digitized as "0"

and "1". Creating an instruction (prompt) for

situations with a larger number of responses (e.g.,

0.0, 0.5, 1.0) may be a task for future research.

Multiple refinements led to the following new

instruction (prompt) for chatGPT 3.5:

"The sample of numbers consists of N=Х

digits, of which N1=Х1 digits "1", N2=Х2 digits

"0" ::

1) Calculate the value of the sample mean M(x).

Write down the value of M(x) ::

2) Calculate Yi as the difference between each

element of the sample and the average value of the

sample, Yi = xi−M(x). Write the values of Yi in the

column ::

3) Calculate Bi as the squares of the differences

between each element of the sample and the average

value of the sample, Bi = (xi−M(x))2. Write down

the Bi values in the column ::

4) Calculate Zi as the product of Ni and Bi, Zi =

Ni × Bi. Write down the values of Zi in the column

:: 5) Calculate Z as the sum of Z = ∑ Zi. Write

down the value of Z ::

6) Calculate C as a quotient of C = Z/N. Write

down the value of C ::

7) Calculate the value of the standard deviation

for the sample δx as the square root of C, that is, δx

= √ C. Write down the value of δx ::

8) Write down the result in the form: M(x) = the

result of calculations up to the fourth decimal place,

δx = the result of calculations up to the fourth

decimal place. Write down the letters M(x) and δx

in bold, please".

This instruction (prompt) contains the following

input symbols, [33], [34]:

- N - the sample size,

- N1, N2 - the numbers of alternative answers (for

WSEAS TRANSACTIONS on COMPUTER RESEARCH

DOI: 10.37394/232018.2024.12.30

Valery Okulich-Kazarin

E-ISSN: 2415-1521

310

Volume 12, 2024

example, for "yes" and for "no").

This instruction (prompt) contains the following

numerical input values:

- Х, Х1, Х2 - the number of respondents'

responses, and Х = Х1 + Х2.

This instruction (prompt) contains the following

output symbols, [33], [34]:

- M(x) - the expected value,

- δx - The standard deviation for the sample.

The rest of the symbols do not matter to the user.

The division of the prompt into separate steps is

caused by a problem that the author encountered in

trying to obtain a result by setting the task of

directly calculating statistical indicators. A research

problem was that chatGPT 3.5 easily performed

computations of statistical indicators for sample

sizes of less than 10 units. However, for larger

sample sizes, chatGPT 3.5 did not yield stable

results. Nevertheless, student graduation projects

require larger sample sizes than 10 units. СhatGPT

3.5 did a good job of dividing the process of

calculating statistical indicators into small

individual steps. Each stage is responsible for

performing one simple mathematical operation.

At the first step, chatGPT calculated the

mathematical expectation M(x). Steps 2-7 were

needed for step-by-step calculation of the standard

deviation for the sample δx. Step 8 gives the

command to print the obtained results with a

precision of 4 decimal places.

A very important advantage of the new chatGPT

3.5 instruction (prompt) lies in relieving the student

from entering data arrays into the new chatGPT 3.5

instructions (prompt).

3.3 The User's Guide to the New chatGPT

Instructions (Prompt)

The user of the new chatGPT 3.5 instructions

(prompt) performs the following actions:

1. Registers on the website

https://chat.openai.com/auth/login.

2. Copy the prompt from section 3.2 of this paper.

3. Input three individual numbers into the

instruction (prompt): X, X1, X2. These

represent the sample size and the number of

responses from respondents for each of the two

alternatives.

4. Pastes the instruction (prompt) into the chatGPT

dialog box and presses the "Send message"

button.

5. Write down the obtained values of the statistical

indicators M(x) and δx. These data will be

calculated with four decimal places.

The advantage of the new instruction (prompt)

is the absence of the need to input arrays of

empirical data or perform any manipulations with

them.

3.4 Five Examples of Calculations of

Statistical Indicators

Five examples are provided in Table 2 to assess the

accuracy of the new instruction (prompt). The initial

data N are taken from Table 1, which presents

sample sizes obtained through mathematical

modeling for various population sizes. Calculations

of statistical indicators M(x) and δx were performed

using three methods:

- Using the new instruction (prompt) for chatGPT

3.5,

- Using Excel tables,

- Using the Windows calculator.

This can be considered one of the limitations of

our study, that two simple and accessible tools were

used to calculate statistical indicators in the

empirical part of the study. We are talking about

Excel tables and the Windows calculator. However,

these are tools accessible to ordinary students and

the use of more complex tools is not planned.

Example 1 combines rows 1-3. Example 2

combines rows 4-6. Example 3 combines rows 7-9.

Example 4 combines rows 10-12. Example 5

combines rows 13-15.

When deciding on the sample size, the

following conditions were chosen when considering

the examples:

- Sampling error 4.00%, [31],

- Standard testing level 0.95, [31],

- Boundaries of the general population from 200 to

50,000.

Intermediate values of the general population size

were selected randomly as psychologically easily

perceived values.

The sample size values N, N1, and N2 for

Examples 1 and 5 (Table 2) were taken from a real-

life case, [35]. They are borrowed from source, [35]

because the sample sizes of N = 144 and N = 599 in

the real cases are close to the modelled sample size

(Table 1), [32]. In the first example, N is slightly

smaller than what was obtained through

mathematical modeling (Table 1). In the fifth

example, N is slightly larger than what was obtained

through mathematical modeling (Table 1).

In Examples 2, 3, and 4 (Table 2), the values of

N1 and N2 were arbitrarily chosen.

WSEAS TRANSACTIONS on COMPUTER RESEARCH

DOI: 10.37394/232018.2024.12.30

Valery Okulich-Kazarin

E-ISSN: 2415-1521

311

Volume 12, 2024

Table 2. Five examples of calculations of statistical

indicators M(x) and δx

№

Тhe tool

M(x)

δx

144

New prompt

0.3542

0.4787

Excel tables

0.3542

0.4783

Windows

0.3542

0.4783

273

240

New prompt

0.1209

0.3269

Excel tables

0.1209

0.3266

Windows

0.1254

0.3302

375

341

New prompt

0.0907

0.2871

Excel tables

0.0907

0.2871

Windows

0.0933

0.2909

500

460

New prompt

0.0800

0.2716

Excel tables

0.0800

0.2713

Windows

0.0820

0.2744

599

534

New prompt

0.1085

0.3114

Excel tables

0.1085

0.3110

Windows

0.1085

0.3110

Table 2 demonstrates a complete match of the

statistical indicators M(x) and δx for all calculation

methods in Examples 1 and 5. In Examples 2, 3, and

4 (Table 2), the values of M(x) and δx coincide for

the new instruction (prompt) and Excel tables.

The figures obtained in Table 2 indicate

differences in the calculation of statistical indicators

using Excel tables and the Windows calculator. The

coincidence of the results of calculations using

Excel tables and the new prompt in all cases

eliminates errors of the new method. The difference

in values obtained using Excel tables and the

Windows calculator could cast doubt on the

reliability of using the Windows calculator.

An example of computing statistical indicators

M(x) and δx using chatGPT 3.5 instructions

(prompt) for a sample with N=375 is presented in

Figure 4 and Figure 5. In Table 2, this example is

located in row 7.

Fig. 4: Screenshot: Example of the new chatGPT 3.5

instruction (prompt) for the sample with N=375

(Table 2, example 3, lines 7-9)

Figure 4 illustrates the detailed appearance of the

instruction (prompt).

Fig. 5: Screenshot: Results of computing statistical

indicators M(x) and δx for the sample with N=375

(Table 2, example 3, lines 7-9)

Figure 5 displays the obtained result in the red

rectangle.

The words "ChatGPT 3.5" have been shifted to

the right by the author in Figure 4 and Figure 5. The

purpose of this shifting is to achieve the compact

figures.

The values of M(x) obtained using ChatGPT

differ by 0.0020-0.0045 from those obtained using

the Windows calculator (Table 2). Thus, the

presented calculation results have a difference of the

third decimal place. This means that high accuracy

of calculations has been achieved, which is

sufficient to complete students’ graduation projects.

Modern prompt engineering has already

accepted the Rule that 3 checks of a new prompt are

a sufficient condition for debugging it

(https://clockwork-

school.com/aistaff?utm_source=vebinar&utm_medi

um=030424&utm_campaign=workshop#about). In

our case, we checked the reliability of the prompt 5

times.

Although you have equal reason to believe in

the accuracy of all three tools, the author

sequentially verifies three pairs of statistical

hypotheses for Examples 2-4.

3.5 Z-statistics

During the verification of the statistical hypotheses,

the author utilized the method of comparing the

means of two independent samples. The essence of

this method lies in computing the Z-statistic, [33],

[34].

Z-statistics are often used to analyze whether

the means of two sets of data are the same, provided

that the population variance is known

(https://habr.com/ru/companies/otus/articles/793678

/). Thus, hypothesis testing is based on assessing the

WSEAS TRANSACTIONS on COMPUTER RESEARCH

DOI: 10.37394/232018.2024.12.30

Valery Okulich-Kazarin

E-ISSN: 2415-1521

312

Volume 12, 2024

significance of the difference M(x1) – M(x2).

When performing the Z-statistic, the research

hypothesis is the assumption that there is no

significant difference between the variables under

study, [33], [34]. In the context of the Z statistic,

which is often used to compare means, research

hypothesis typically states that the means of the two

groups being compared are equal. The alternative

hypothesis is the opposite of the null hypothesis: the

means of the two groups being compared are not

equal. Therefore, the author chose the Z-statistics

method.

The comparison of the two means will be

conducted for two methods of calculating statistical

indicators: using the new prompt and using the

Windows calculator. The comparison is carried out

for three pairs of rows in Table 2:

- Row 4 and Row 6,

- Row 7 and Row 9,

- Row 10 and Row 12.

So, the author made three pairs of hypotheses,

[33], [34]: a Research hypothesis and an Alternative

hypothesis.

Research hypothesis: M(х1) – M(х2) = 0.00.

That is, if you do not take into account random

deviations.

Alternative hypothesis: M(х1) – M(х2) ≠ 0.00.

That is, if you do not take into account random

deviations. In this case, the difference can be either

greater than 0.00% or less than 0.00%.

If the difference M(x1) – M(x2) equals 0.00, it

means that the two samples are equal to each other.

There is no statistically significant difference

between the two means. The results of calculating

the statistical indicator M(x) using two methods are

equal – their difference is explained by random

deviations.

If the difference M(x1) – M(x2) is greater or

less than 0.00, it means that the two samples are not

equal to each other. There is a statistically

significant difference between the two means. The

results of calculating the statistical indicator M(x)

using two methods differ from each other.

A detailed description of the Z-statistic is

provided in statistics textbooks, including [33], [34],

[36]. Here, a two-sided test was adopted because the

comparison result can be both greater than 0.00%

and less than 0.00%.

For each verification case (Table 3, Table 4,

Table 5), the author used the standard significance

level, [33], which is 0.05 (α = 0.05).

Table 3. Results of comparing two sample averages

for row 4 and row 6

Calculation stage

Row 4

Row 6

The size of a sample, N

273

The expected value, M(x), %

0.1209

0.1254

| M(x1) – M(x2) |

0.0045

μ1 – μ2

0.00

The standard deviation for

the sample, δх

0.3269

0.3302

Average error, ṠẊ = δх / √n

0.0198

0.0200

ṠẊ2

0.00039

0.000340

| Ṡ12 - Ṡ22 |

0.00001

√ (Ṡ12 - Ṡ22)

0.00316

| zstat | = [M(x1) – M(x2) - (μ1 – μ2)]

/ √ (Ṡ12 - Ṡ22)

1.4240

The value ztabl for the standard testing

level of 0.05 [33,34]

1.96

Result, | zstat | > ztabl

Table 3 shows that the Z-statistics | zstat | is less

than the ztabl. In this case, the Research hypothesis is

accepted: M(x1) – M(x2) = 0.00. This means that

the results of calculating of statistical indicator M(x)

are equal for both ways – their difference is

explained by random deviations.

Table 4. Results of comparing two sample averages

for row 7 and row 9

Calculation stage

Row 7

Row 9

The size of a sample, N

375

The expected value, M(x), %

0.0907

0.0933

| M(x1) – M(x2) |

0.0026

μ1 – μ2

0.00

The standard deviation for

the sample, δх

0.2871

0.2909

Average error, ṠẊ = δх / √n

0.0148

0.0150

ṠẊ2

0.00022

0.00023

| Ṡ12 - Ṡ22 |

0.00001

√ (Ṡ12 - Ṡ22)

0.00316

| zstat | = [M(x1) – M(x2) - (μ1 – μ2)]

/ √ (Ṡ12 - Ṡ22)

0.8228

The value ztabl for the standard testing

level of 0.05 [33,34]

1.96

Result, | zstat | > ztabl

Table 4 shows that the Z-statistics | zstat | is less

than the ztabl. In this case, the Research hypothesis is

accepted: M(x1) – M(x2) = 0.00. This means that

the results of calculating of statistical indicator M(x)

are equal for both ways – their difference is

explained by random deviations.

WSEAS TRANSACTIONS on COMPUTER RESEARCH

DOI: 10.37394/232018.2024.12.30

Valery Okulich-Kazarin

E-ISSN: 2415-1521

313

Volume 12, 2024

Table 5. Results of comparing two sample averages

for row 10 and row 12

Calculation stage

Row 10

Row 12

The size of a sample, N

500

The expected value, M(x), %

0.0800

0.0820

| M(x1) – M(x2) |

0.0020

μ1 – μ2

0.00

The standard deviation for

the sample, δх

0.2716

0.2744

Average error, ṠẊ = δх / √n

0.0121

0.0123

ṠẊ2

0.000147

0.000151

| Ṡ12 - Ṡ22 |

0.000004

√ (Ṡ12 - Ṡ22)

0.002

| zstat | = [M(x1) – M(x2) –

(μ1 – μ2)] / √ (Ṡ12 - Ṡ22)

1.000

The value ztabl for the standard

testing level of 0.05 [33,34]

1.96

Result, | zstat | > ztabl

Table 5 shows that the Z-statistics | zstat | is less

than the ztabl. In this case, the Research hypothesis is

accepted: M(x1) – M(x2) = 0.00. This means that

the results of calculating of statistical indicator M(x)

are equal for both ways – their difference is

explained by random deviations.

Thus, the Z-statistic confirmed the equality of

the statistical indicators calculated using the new

chatGPT 3.5 instruction (prompt) and two other

traditional methods. The standard testing level is α =

0.05.

This means that students can use the new

chatGPT 3.5 instruction (prompt) to calculate

statistical indicators when completing graduation

and other projects. However, we may encounter

both a prejudice that the use of artificial intelligence

is not in demand for simple statistical calculations,

and a prejudice about the uselessness of Artificial

Intelligence in general.

4 Conclusion

The purpose of the study has been achieved.

The research problem has two main aspects.

Firstly, the three stages of bibliometric analysis on

79606 sources published in the Scopus database

showed increased interest in addressing issues

related to "graduation project" and "statistical

indicators." The research results suggest the need for

finding a way to calculate statistical indicators

without manual computations. Secondly, the

possibility of errors in calculations using chatGPT

3.5 was empirically confirmed, as well as in

scientific literature.

The novelty of the work lies in the creation of

the new chatGPT 3.5 instruction (prompt) to

calculate statistical indicators for student graduation

projects. The new chatGPT 3.5 instruction (prompt)

enables the calculation of statistical indicators

without manual computations. The significant

advantage of the new chatGPT 3.5 instruction

(prompt) lies in relieving the student from entering

data arrays into the tool for computing statistical

indicators.

The created new chatGPT 3.5 instruction

(prompt) should become a valuable resource for

students, helping them save time in preparing their

graduation projects. Students, instructors, and other

interested parties can use the prompt from section

3.3.

The study has scientific and practical

significance, since it reduces the time and simplifies

the process of calculating statistical indicators of

graduation projects, and also improves their quality:

1. New ChatGPT 3.5 instruction (prompt) is a

useful resource for students to save time on

graduation projects.

2. Using the proposal allows you to improve the

quality of graduation projects.

3. The new ChatGPT 3.5 instruction (prompt) can

help students with various statistics knowledge.

4. This discovery has the potential to expand the

use of computers in education.

One of the limitations of the study is the choice

of the simplest option for calculating statistical

indicators. As empirical data and prompt

engineering experience are collected, instructions

(prompts) will be compiled for solving more

complex statistical problems. Exploring how the

performance of ChatGPT 3.5 might vary with

different types of statistical indicators, sample sizes,

or data distributions is one of the next steps of our

study. The next goal of the research is to create a

new chatGPT 3.5 instruction (prompt) for

computing statistical indicators for a larger number

of respondents (units). Creating instructions

(prompts) for situations with a larger number of

responses (e.g., 0.0, 0.5, 1.0) and receiving quality

feedback from students who use the new instruction

may also be the tasks for future research.

Acknowledgement:

The study was carried out with the support of the

Eastern European Scientific Group. The author

would like to thank the reviewers for their valuable

comments.

WSEAS TRANSACTIONS on COMPUTER RESEARCH

DOI: 10.37394/232018.2024.12.30

Valery Okulich-Kazarin

E-ISSN: 2415-1521

314

Volume 12, 2024

References:

[1] Sassy Bayona-Oré, Student Learning Styles in

Information Technology, WSEAS

Transactions on Computer Research, vol. 11,

2023, pp. 385-392,

https://doi.org/10.37394/232018.2023.11.35.

[2] Razeale G. Resultay, Rudjane C. Tunac, Mary

Joy O. Macaraeg, Anti-Bullying Intervention

(ABBI) Program Manual: Grounding

Teachers and Counselors, WSEAS

Transactions on Computer Research, vol. 12,

2024, pp. 181-195,

https://doi.org/10.37394/232018.2024.12.18.

[3] Jakoet-Salie, Amina, Ramalobe, Kutu, The

digitalization of learning and teaching

practices in higher education institutions

during the Covid-19 pandemic, Teaching

Public Administration, Vol. 41, No.1, 2023,

pp.59-71. DOI: 10.1177/01447394221092275.

[4] Mariam A. Shomoye, Najeem O. Adelakun,

Kehinde L. Adebisi, Exploring the Use of

Electronic Resources for Undergraduate

Learning at the National Open University of

Nigeria in Kwara State, WSEAS Transactions

on Computers, vol. 22, 2023, pp. 302-310,

https://doi.org/10.37394/23205.2023.22.34.

[5] Siti Nur Zahirah Omar, Rusnifaezah Musa,

Maliani Mohamad, Che Mohd Syaharuddin

Che Cob, Azmahani Yaacob Othman, Razli

Ramli, Efficiency Online Learning During

Covid-19 Pandemic, WSEAS Transactions on

Business and Economics, Vol. 20, 2023, pp.

30-39,

https://doi.org/10.37394/23207.2023.20.4.

[6] Liliana Sobreira, David Pascoal Rosado,

Helga Santa Comba Lopes, Maria Jose Sousa,

The Implementation of e-Learning in the

Continuous Training of the Military of the

Republican National Guard. Case Study:

Lisbon Territorial Command, WSEAS

Transactions on Systems and Control, Vol.

16, 2021, pp. 66-76,

https://doi.org/10.37394/23203.2021.16.4.

[7] Nursel Selver Ruzgar, Clare Chua-Chow,

Perception of Students on Online Exams and

How Sequential Exams and the Lockdown

Browser Affect Student Anxiety and

Performance, WSEAS Transactions on

Computer Research, vol. 11, 2023, pp. 92-

110,

https://doi.org/10.37394/232018.2023.11.9.

[8] Bazzi A., Chafii M., On Outage-Based

Beamforming Design for Dual-Functional

Radar-Communication 6G Systems, IEEE

Transactions on Wireless Communications,

vol. 22, no. 8, 2023, pp. 5598-5612, doi:

10.1109/TWC.2023.3235617.

[9] Bazzi A., Chafii M., Secure Full Duplex

Integrated Sensing and Communications,

IEEE Transactions on Information Forensics

and Security, vol. 19, 2024, pp. 2082-2097,

doi: 10.1109/TIFS.2023.3346696.

[10] Pence H., Artificial Intelligence in Higher

Education: New Wine in Old Wineskins?

Journal of Educational Technology Systems,

Vol.48, No.1, 2019, pp. 5-13. doi:

10.1177/0047239519865577.

[11] Gellai Dániel Béla, Enterprising Academcs:

Heterarchical Policy Networks for Artificial

Intelligence in British Higher Education,

ECNU Review of Education, December 19,

2022. doi: 10.1177/20965311221143798.

[12] West C. G., AI and the FCI: Can ChatGPT

project an understanding of introductory

physics? 2023, arXiv preprint

arXiv:2303.01067

[13] Choi J., Hickman K., Monahan A., Schwarcz

D., ChatGPT goes to law school. Minnesota

Legal Studies Research Paper, 23(03), 2023.

[14] Leswing K., OpenAI announces GPT-4,

claims it can beat 90% of humans on the SAT.

CNBC, [Online].

https://www.cnbc.com/2023/03/14/openai-

announces-gpt-4-says-beats-90percent-of-

humans-on-sat.html (Accessed Date: March

26, 2023).

[15] Geerling Wayne, Mateer G. Dirk, Wooten

Jadrian, Damodaran Nikhil. ChatGPT has

Aced the Test of Understanding in College

Economics: Now What? The American

Economist, 68(2), 2023, рр. 233-245, DOI:

10.1177/05694345231169654.

[16] Gilson A., Safranek C. W., Huang T.,

Socrates V., Chi L., Taylor R. A., Chartash D.

How does ChatGPT perform on the United

States medical licensing examination? The

implications of large language models for

medical education and knowledge assessment,

JMIR Medical Education, Vol.9, No.1,

e45312, 2023. DOI: 10.2196/45312.

[17] Wilson S. G. The Flipped Class: A Method to

Address the Challenges of an Undergraduate

Statistics Course, Teaching of Psychology,

40(3), 2013, рр. 193-199, DOI:

10.1177/0098628313487461.

[18] Trinh M. P., Chico R. J., Reed R. M. How

Fun Overcame Fear: The Gamification of a

Graduate-Level Statistics Course, Journal of

Management Education, 0(0), 2023,

https://doi.org/10.1177/10525629231181120.

WSEAS TRANSACTIONS on COMPUTER RESEARCH

DOI: 10.37394/232018.2024.12.30

Valery Okulich-Kazarin

E-ISSN: 2415-1521

315

Volume 12, 2024

[19] Pirlott A. G., Hines J. C. Eliminating ANOVA

Hand Calculations Predicts Improved Mastery

in an Undergraduate Statistics Course.

Teaching of Psychology, 0(0), 2023, DOI:

10.1177/00986283231183959.

[20] Reynolds J. J. Let’s Talk about Stats:

Revising Our Approach to Teaching Statistics

in Psychology. Psychological Reports, 126(1),

2023, рр. 5-33, DOI:

10.1177/00332941211043447.

[21] Callingham R., Watson J., Oates G. Learning

progressions and the Australian curriculum

mathematics: The case of statistics and

probability, Australian Journal of Education,

65(3), 2021, рр. 329-342,

https://doi.org/10.1177/00049441211036521.

[22] Eastridge J. A., Benson W. L. Comparing

Two Models of Collaborative Testing for

Teaching Statistics, Teaching of Psychology,

47(1), 2020, рр. 68-73,

https://doi.org/10.1177/0098628319888113.

[23] Paltoglou A. E., Morys-Carter W. L., Davies

E. L. From Anxiety to Confidence: Exploring

the Measurement of Statistics Confidence and

its Relationship with Experience, Knowledge

and Competence within Psychology

Undergraduate Students, Psychology Learning

& Teaching, 18(2), 2019, рр. 165-178,

https://doi.org/10.1177/1475725718819290.

[24] Fielding S, Poobalan A, Prescott G, Marais D,

Aucott L. Views of medical students: what,

when and how do they want statistics taught?

Scottish Medical Journal, 60(4), 2015, рр.

164-169. doi:10.1177/0036933015608329.

[25] Ralston K. Sociologists Shouldn’t Have to

Study Statistics: Epistemology and Anxiety of

Statistics in Sociology Students. Sociological

Research Online, 25(2), 2020, рр. 219-235.

https://doi.org/10.1177/1360780419888927

[26] Walker R. V., Osborn, H., Madden, J., &

Jennings Black, K. Integrating Diversity into

Psychology Statistics Courses: Advice,

Reflections, and Special Considerations.

Teaching of Psychology, 0(0), 2023,

https://doi.org/10.1177/00986283231199461.

[27] Yang H.-H., Lin Y.-T. How Knowledge

Sharing and Cohesion Become Keys to a

Successful Graduation Project for Students

from Design College, SAGE Open, 12(3),

2022,

https://doi.org/10.1177/21582440221121785.

[28] Conner J. O. Student Engagement in an

Independent Research Project: The Influence

of Cohort Culture, Journal of Advanced

Academics, 21(1), 2009, рр. 8-38.

https://doi.org/10.1177/1932202X0902100102

[29] Herzallah R. Practically Oriented Design

Projects in Mechatronics Engineering: A Case

Study of Design Experiences and Outcomes,

International Journal of Electrical

Engineering & Education, 47(1), 2010, рр.

31-46. doi:10.7227/IJEEE.47.1.4.

[30] Leswing K., OpenAI announces GPT-4,

claims it can beat 90% of humans on the SAT.

CNBC, [Online].

https://www.cnbc.com/2023/03/14/openai-

announces-gpt-4-says-beats-90percent-of-

humans-on-sat.html (Accessed Date: March

26, 2023).

[31] Sample. Size is not the main thing. Or the

main thing, [Online].

https://scanmarket.ru/blog/vyborka-razmer-

ne-glavnoe-ili-glavnoe (Accessed Date:

January 31, 2024).

[32] A calculator to calculate a sufficient sample

size, [Online].

https://scanmarket.ru/blog/vyborka-razmer-

ne-glavnoe-ili-glavnoe#calc1 (Accessed Date:

May 24, 2024).

[33] Singpurwalla D., A Handbook of Statistics: An

Overview of Statistical Methods, bookboon.

2015.

[34] Kravchenko A., Sociologia: textbook for

students [in Russian], Yurayt. 2014.

[35] Okulich-Kazarin V., Artyukhov A., Skowron

Ł., Artyukhova N., Dluhopolskyi O., Cwynar

W. Sustainability of Higher Education: Study

of Student Opinions about the Possibility of

Replacing Teachers with AI

Technologies, Sustainability, 16(1), 2024, 55,

https://doi.org/10.3390/su16010055.

[36] BUS_9641. Business_Statistics_5, Textbook

for the Program 'Masters of Business

Administration', Kingston University, 2010.

WSEAS TRANSACTIONS on COMPUTER RESEARCH

DOI: 10.37394/232018.2024.12.30

Valery Okulich-Kazarin

E-ISSN: 2415-1521

316

Volume 12, 2024

Contribution of Individual Authors to the

Creation of a Scientific Article (Ghostwriting

Policy)

The author contributed in the present research, at all

stages from the formulation of the problem to the

final findings and solution.

Sources of Funding for Research Presented in a

Scientific Article or Scientific Article Itself

No funding was received for conducting this study.

Conflict of Interest

The authors have no conflicts of interest to declare.

Creative Commons Attribution License 4.0

(Attribution 4.0 International, CC BY 4.0)

This article is published under the terms of the

Creative Commons Attribution License 4.0

https://creativecommons.org/licenses/by/4.0/deed.en

_US

WSEAS TRANSACTIONS on COMPUTER RESEARCH

DOI: 10.37394/232018.2024.12.30

Valery Okulich-Kazarin

E-ISSN: 2415-1521

317

Volume 12, 2024