Using the Ordinal Probit regression model to study the addiction of young people to the game (PUBG) and its impact on family cohesion.

doi:10.21203/rs.3.rs-3576349/v1

Download PDF

Research Article

Using the Ordinal Probit regression model to study the addiction of young people to the game (PUBG) and its impact on family cohesion.

https://doi.org/10.21203/rs.3.rs-3576349/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

This research is based on a set of main objectives, which are to identify the most influential factors on family cohesion using the Ordinal Probit regression model and to determine which independent variables positively contribute to the dependent variable. Additionally, the study aims to test hypotheses that suggest that the Ordinal Probit regression model is the most appropriate statistical method for analyzing study data, and that certain variables related to young people addicted to the video game ‘PUBG’ (gender, age, education level, social status, occupation, type device to play, time of entry into the game, number of hours of play, method of play, Motivation for using the game, and places of play) have an impact on family cohesion. The study was conducted with a confidence level of 95% and a sample size of 384 individuals, and its type is simple random sampling. The research found a number of results, including that the Probit model is a suitable method for the data. It also showed that variables (gender, age, social status, favorite device to play, number of hours played, and motive for play) have the most significant impact on family cohesion and are important in interpreting the dependent variable (family cohesion). The highest contribution was found to be from the number of hours of play with 49.75%, followed by age at 12.02%, then the motives for play at 9.94%, followed by social status, gender, and method of play consecutively.

Ordinal Probit Regression، Test of Parallel Lines، Model Fitting Information، Pseudo R-Square

Video games have invaded the lives of children and young people; even it became a necessity in their daily practices. With technological advancements and improvements in graphics and sound quality, video games have been able to attract young people to it. In the recent years, a game called PUBG has appeared and the attraction of young people to it has been significant. It is rare to find a family without a PUBG player, which has had an impact on many aspects of their social, family, and behavioral lives. Therefore, it is necessary to study it scientifically, and the Ordinal Probit regression is considered of the predictive statistical tests that allow researchers to predict the effect of some variables on an ordinal dependent variable. This gives the research significant scientific importance.

Problem of the Research:

This research uses the ordinal logistic regression model to study the impact of PUBG on family cohesion, considering that the family is the first nucleus of the individual and the closest center to him and is the first to be affected by his behaviors. Therefore, it was necessary to study and analyze the effects of PUBG on family cohesion. Thus, the topic of this research is applying the ordinal Probit regression model to study the youth’s addiction to PUBG and its impact on family cohesion.

Importance of the Research:

The importance of this research lies in its ability to identify the effects that result from playing PUBG on family cohesion, and attempting to benefit from its results by predicting the variables that contribute to the impact on family cohesion. This can contribute to the development by using its results in setting developmental plans. The importance of this research also lies in the fact that the results that can be reached may be of practical and scientific benefit, and its recommendations can be utilized in the field of family.

Research Objectives:

Identifying the most influential factors on family cohesion by using the ordinal Probit regression model and reaching results that can be relied upon to predict family cohesion.
Knowing whether the independent variables positively contribute to the dependent variable or not.
Building a statistical model that shows the relationship between youth addiction to PUBG and family cohesion.

Research Hypotheses:

1. The ordinal Probit regression model is the most appropriate statistical method for analyzing the study's data.

2. Some variables related to addicted youth to PUBG (gender, age, education level, social status, occupation, type of device, time of entry into the game, number of hours of play, method of play, motives for use, places of use) affect family cohesion.

Research Community and Sample

The research community is all Libyan citizens, whose number was approximately 6,103,100 (According to the 2015 statistics). Due to the large amount of the community, we used the sampling method and relied on Steven Thompson's equation to evaluate the sample size ratio in the case of a limited community, assuming that the error rate is 5% and the confidence level is 95%. The sample size was 384 individuals, and it was a simple random sample.

Table (1): Research Variables:

Variable	Code	Categories (Coding)
Family cohesion (The dependent variable)	y	1 = Weak cohesion 2 = Medium cohesion 3 = Strong cohesion
Gender	x₁	1 = Male 2 = Female
Age	x₂	1 = From 15 to 22 years 2 = From 23 to 30 years 3 = From 31 to 38 years
Education level	x₃	1 = Secondary or lower 2 = University or higher
Social status	x₄	1 = Single 2 = Married 3 = Divorced 4 = Widowed
Occupation	x₅	1 = Don’t work 2 = Student 3 = Government Employee 4 = Private Business
Favorite device to play	x₆	1 = Phone 2 = Computer (Or laptop) 3 = Tablets
Time of entry into the game	x₇	1 = Daily 2 = Weekly 3 = Monthly
Number of hours of play	x₈	1 = Less than 2 hours 2 = From 2 to less than 5 hours 3 = From 5 to less than 8 hours 4 = From 8 hours or more
Method of play	x₉	1 = Playing more than one round continuously 2 = Playing intermittently and taking a break between rounds
Motivation for using the game	x₁₀	1 = Entertainment and leisure 2 = Spending free time 3 = Making new friends 4 = To release anger and frustration 5 = Developing a skill 6 = Developing a talent
Places of play	x₁₁	1 = Home 2 = With friends 3 = At work 4 = All of the above 5 = I don’t have a specific place

Table (2): General properties (Frequencies and Proportions) for the research Sample

Percent	Frequency	Categories	Variable
65.6%	252	Male	Gender
%34.4	132	Female	Gender
50.5%	194	From 15 to 22 years	Age
34.9%	134	From 23 to 30 years
14.6%	56	From 31 to 38 years
31.0%	119	Secondary or lower	Education level
69.0%	265	University or higher	Education level
76.3%	293	Single	Social status
19.0%	73	Married
2.9%	11	Divorced
1.8%	7	Widowed
10.4%	40	Don’t work	Occupation
62.0%	238	Student
14.8%	57	Government Employee
12.8%	49	Private Business

From the table (2) it appeared to us that the number of males in the sample is higher than the number of females, with a percentage of 65.6%. Most of them were in the age group of 15 to 22 years, with a percentage of 50.5%. The educational level of most of them was university or higher, with a percentage of 69%. As for the marital status, the majority of them were single, with a percentage of 76.3%. Regarding occupation, 62% of the sample were students.

Table (3): Data related to the research participants interaction with PUBG

Percent	Frequency	Categories	Variable
81.5%	313	Phone	Favorite device to play
7.8%	30	Computer (Or laptop)
10.7%	41	Tablets
59.4%	228	Daily	Time of entry into the game
26.3%	101	Weekly
14.3%	55	Monthly
35.7%	137	Less than 2 hours	Number of hours of play
37.2%	143	From 2 to less than 5 hours
%16.9	65	From 5 to less than 8 hours
10.2%	39	From 8 hours or more
54.7%	210	Playing more than one round continuously	Method of play
45.3%	174	Playing intermittently and taking a break between rounds	Method of play
52.6%	202	Entertainment and leisure	Motivation for using the game
22.7%	87	Spending free time
6.3%	24	Making new friends
7.8%	30	To release anger and frustration
4.9%	19	Developing a skill
5.7%	22	Developing a talent
59.4%	228	Home	Places of play
8.6%	33	With friends
1.3%	5	At work
8.3%	32	All of the above
22.4%	86	I don’t have a specific place

From the table, it appeared to us that the preferred device for PUBG players is the phone, with a percentage of 81.5%. As for their entry of playing, the highest percentage of the sample, 59.4%, play on a daily basis. The highest percentage of players spend from 2 to less than 5 hours playing, with a percentage of 37.2%. Additionally, 54.7% of them play the game for more than one round continuously. The majority of players, with a percentage of 52.6%, view the motivation behind playing the game as entertainment and leisure. The home is the most common place for playing the game, with a percentage of 59.4% of the sample.

Ordinal Probit regression

The Ordinal regression model was built using the (Probit) method for each variable of the study, considering that the dependent variable is a categorical ordinal variable, as follows:

Table (4): Dependent Variable

family cohesion
Internal Value	Original Value
1	Weak family cohesion
2	Medium family cohesion
3	Strong family cohesion

Results of the Multicollinearity Assumption between Independent Variables

Before initiating the ordinal regression analysis, it is necessary to check for the absence of multicollinearity among the independent variables included in the study, using the variance inflation factor (VIF). The VIF values were as follows:

Table (5): variance inflation factor

VIF	Variable
1.186	x₁	1.
1.259	x₂	2.
1.177	x₃	3.
1.161	x₄	4.
1.215	x₅	5.
1.089	x₆	6.
1.413	x₇	7.
1.584	x₈	8.
1.175	x₉	9.
1.097	x₁₀	10.
1.060	x₁₁	11.

Based on the results, we found that the VIF values range from 1.060 to 1.584, and all of these values are less than 3. This indicates the absence of multicollinearity among the variables.

Test of Parallel Lines

Before interpreting the estimated coefficients of the ordinal regression model, it is necessary to test the assumption of equality of regression coefficients, which is an important performance measure to ensure the quality of the model fit.

Table (6): Test of Parallel Lines

Test of Parallel Lines^a
Model	-2 Log Likelihood	Chi-Square	df	Sig.
Null Hypothesis	415.945
General	400.987	14.959	11	0.184
The null hypothesis states that the location parameters (slope coefficients) are the same across response categories.
a. Link function: Probit.

In these data, we observe that the value of (sig = 0.184), which is greater than 0.05. This indicates that the null hypothesis is accepted, meaning that the model is validated. Based on the results of the table, we can use the Ordinal Probit regression.

Model Fitting Information Test

Compares the model in the case of no independent variables (with only the intercept) and in the case of having independent variables. The Model Fitting Information table of the analysis gives the − 2 log likelihood values for the model established without independent variables and that with independent variables as follows:

Table (7): Model Fitting Information

Model Fitting Information
Model	-2 Log Likelihood	Chi-Square	df	Sig.
Intercept Only	575.062
Final	415.945	159.117	11	0.000
Link function: Probit.

From the previous table, we observe that:

(${H}_{0}$): (The model is not significant), meaning that the independent variables do not contribute positively to the dependent variable.

(${H}_{1}$): (The model is significant), meaning that the independent variables contribute positively to the dependent variable.

In these data, we observe that the significance value (sig = 0.00) is less than 0.05. This means that we reject the null hypothesis, indicating that the model is significant (i.e., the independent variables contribute positively) both with the constant term and with the independent variables. and this indicates the existence of a relationship between the dependent variable and the independent.

Table (8): Model of Fit Test

Goodness-of-Fit
	Chi-Square	df	Sig.
Pearson	657.112	633	0. 246
Deviance	397.161	633	1.000
Link function: Probit.

(${H}_{0}$): Probit model fits the data (the difference is not significant).

(${H}_{1}$): Probit model does not fit the data (the difference is significant).

This test relies on the significance value (sig-Deviance). In these data, we observe that the significance value (sig) is greater than 0.05, which means that we accept the null hypothesis. This indicates that the Probit model fits the data, and the difference between the expected and observed values is not significant.

Pseudo R-Square Test

The pseudo-R² used to assess the degree of fit of the model. It is a statistical model that describes how well the model fits a set of observations, also The pseudo-R² value aims to measure and assess the power of the relation between the dependent variable and the independent variables. The McFadden، Cox- Snell، and Nagelkerke R² statistics are the most used pseudo-R² statistics.

Table (9): Pseudo R-Square

Pseudo R-Square
Cox and Snell	0.339
Nagelkerke	0.429
McFadden	0.264
Link function: Probit.

The variables included in the model explained about 42.9% using the Nagelkerke R Square coefficient and approximately 33.9% using the Cox & Snell R Square coefficient of the variations in the effect of the dependent variable. This indicates that there is still a proportion of the variations in the dependent variable that can be explained by other variables not included in the model.

Table (10): Expression of the Significances of the Model Parameters

Parameter Estimates
		Estimate	Std. Error	Wald	df	Sig.
Threshold	[y = 1.00]	-2.065	0.577	12.818	1	0.000
Threshold	[y = 2.00]	0.832	0.568	2.146	1	0.143
Location	x₁	-0.321	0.157	4.162	1	0.041
	x₂	-0.380	0.110	12.027	1	0.001
	x₃	0.237	0.160	2.204	1	0.138
	x₄	0.278	0.119	5.439	1	0.020
	x₅	0.151	0.093	2.673	1	0.102
	x₆	-0.213	0.110	3.747	1	0.053
	x₇	0.156	0.111	1.978	1	0.160
	x₈	-0.674	0.095	49.751	1	0.000
	x₉	0.298	0.148	4.079	1	0.043
	x₁₀	-0.151	0.048	9.942	1	0.002
	x₁₁	-0.006	0.041	0.024	1	0.876

From the ‘threshold’ table, we can see that the significance of the first interval value has been achieved, while the significance of the second interval value has not been achieved. This indicates that the explanatory variables studied were not clearly specified for the difference in the strength of the effect between the weak impact option and the moderate impact option.

Additionally, the ‘Estimate’ value shows that any respondent who obtains a value of less than − 2.065 has a weak probability of family cohesion, while those with values between − 2.065 and 0.832 have a medium probability of family cohesion. Those with values greater than 0.832 have a strong probability of family cohesion. Looking at the significant variable values in the ‘Estimate’ column, we notice that they range between − 0.151 and 0.278, all falling within the range of medium family cohesion.

Furthermore, we observe that the significance value (sig) of the variables (gender, age, social status, favorite device to play, hours of play, method of play and motivation for using the game) is less than or equal to 0.05. This means that these variables are important in explaining the dependent variable (family cohesion).

However, the significance value (sig) of the variables (educational level, occupation, time of entry into the game, and places of play) is greater than 0.05. This means that these variables are not significant in explaining the dependent variable (family cohesion)

To determine the contribution percentage of each variable, we look at the ‘Wald’ value. We notice that the highest contribution was for hours of play at 49.75%, followed by age at 12.02%, then the motives for play at 9.94%, and then social status, gender, and method of play in descending order.

$${y}^{*}={\widehat{B}}_{1}{x}_{1}+{\widehat{B}}_{2}{x}_{2}+{\widehat{B}}_{3}{x}_{3}+{\widehat{B}}_{4}{x}_{4}+{\widehat{B}}_{5}{x}_{5}+{\widehat{B}}_{6}{x}_{6}+{\widehat{B}}_{7}{x}_{7}+{\widehat{B}}_{8}{x}_{8}+{\widehat{B}}_{9}{x}_{9}+{\widehat{B}}_{10}{x}_{10}+{\widehat{B}}_{11}{x}_{11}$$

$${y}^{*}=-0.321{x}_{1}-0.380{x}_{2}+0.237{x}_{3}+0.278{x}_{4}+0.151{x}_{5}-0.213{x}_{6}+0.156{x}_{7}-0.674{x}_{8}+0.298{x}_{9}-0.151{x}_{10}-0.006{x}_{11}$$

1. Applying the Probit model to the study data showed that it is a suitable method for the data.

2. The significance of the model indicated that the independent variables have a positive contribution to the dependent variable. The variables included in the model explained approximately 42.9% of the variations in the effect of the dependent variable, using the ‘Nagelkerke R Square’ coefficient.

3. The variables (gender, age, social status, favorite device, hours of play, method of play and motivation for using the game) were found to have the most significant impact on family cohesion and are important in explaining the dependent variable (family cohesion).

4. The highest contribution was found for hours of play at 49.75%, followed by age at 12.02%, then the motivation for using the game at 9.94%. Social status, gender, and method of play had contributions in descending order.

1. Expanding the use of the ordinal Probit regression model in different fields and issues.

2. The necessity for attention from specialized entities to the issue of video games and their impact on family issues, such as through awareness campaigns to clarify the negatives of video games.

3. Holding conferences and seminars that focus on the application of the ordinal Probit regression model to various scientific topics.

4. Conducting researches and studies in various specialties on the topic of video games in general, and the game PUBG in particular, given the scarcity of studies on this topic in the Libyan community as far as we know.

Lei, C., Sun, X.: Comparing lethal dose ratios using probit regression with arbitrary slopes. BMC Pharmacol. Toxicol. 19, 61 (2018)
Bozpolat, E.: Investigation of the self-regulated learning strategies of students from the faculty of education using ordinal logistic regression analysis. Educational Sciences: Theory and Practice. 16(1), 301–318 (2016)
Adejumo, A.O., Adetunji, A.A.: Application of Ordinal Logistic Regression in the Study of Students’ Performance. J. Math. Theory Modelling. 3(11), 10–19 (2013)
Hannah, M., et al.: Presentation of Ordinal Regression Analysis on the Original Scale. Biometrics; pp. 771–775. (1996)
Jackman, S.: Models for Ordered Outcomes. Political Science 200C. Stanford University, v. 150C/350C, p. 1–14,. (2000)
Becker, W.E., Peter, E., Kennedy, “A Graphical Exposition of the Ordered Probit,” Econometric Theory, 8:127–31, Boes, S., Winkelmann, R.: (2006): “Ordered Response Models,” in Modern Econometric Analysis, Springer, Chap. 12, 167–181. (1992)
Xing Liu, H., Koirala: Ordinal Regression Analysis: Using Generalized Ordinal Logistic Regression Models to Estimate Educational Data (2012)
Williams, R.: Ordinal regression models: Problems, solutions, and problems with the solutions. (2008)
Marlena, L., Nugraheni, E.A.: 'Probit Regression Analysis in Estimating the Effect of Learning Assisted by Cabri 3D on Students' Mathematical Understanding Ability'. Jurnal Pendidikan Matematika, Al-Jabar (2019)
Della Lucia, S.M., et al.: '' Ordered probit regression analysis of the effect of brand name on beer acceptance by consumers'' Food Science and Technology. PP 586–591. (2013)

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Using the Ordinal Probit regression model to study the addiction of young people to the game (PUBG) and its impact on family cohesion.

Status:

Version 1

Abstract

Introduction

Problem of the Research:

Importance of the Research:

Research Community and Sample

Ordinal Probit regression

Results of the Multicollinearity Assumption between Independent Variables

Test of Parallel Lines

Model Fitting Information Test

Pseudo R-Square Test

Results

Recommendations

References

Additional Declarations

Status:

Version 1