Application of Generalized Estimation Equation in the Analysis of Factors Affecting the Continuous Use of College English Education Platforms

In order to explore the factors influencing the use of generalized estimating equations in English education platforms, the author proposes a research method for the continuous use of English education platforms. The author put forward 4 hypotheses and 12 scale questions, and designed a questionnaire based on them, taking perceived interaction, content quality, and platform serviceability as potential basic variables, the author constructed a model of influencing factors for college students' continuous use of English education and learning platforms, explore the key factors that affect the sustainable use of college students' English education and learning platform. The results show that the main influencing factors of users' intention to use English education and learning platform are content quality, perceived interaction and platform service, therefore, it is proposed to focus on the content quality, pay attention to the user experience, strengthen the interaction of the platform, and propose reasonable pricing to implement the pricing mechanism, so as to improve college students' intention to continue using the English education and learning platform.


Introduction
The generalized estimation equation is a method to study longitudinal data.There may be a correlation between the results of multiple measurements of the same measurement object.If the correlation between the data is not considered, it will cause information loss.Common research models (such as linear regression) require data to be independent, in this case, generalized estimation equation can be used for research.The continuous measurement of variance requires that the data be complete and not missing, but in actual research, missing data is more common, and the generalized estimation equation can also be used for research at this time.The difference is that the repeated measurement variance is analyzed from the perspective of difference relationship, but the generalized estimation equation is analyzed from the perspective of influence relationship.In addition, repeated measurement variance requires that dependent variable Y is quantitative continuous data and independent variable X is categorical data; However, when estimating the equation in a broad sense, the dependent variable Y is quantitative data or binary data, or Poisson distribution or negative binomial distribution data, there is no special requirement for the data type of the independent variable, if it is classified data, you can directly set virtual variables [1][2].As shown in Figure 1: In order to explore the factors that affect the use of generalized estimation equation in English education platform, this study uses data as longitudinal data, namely repeated measurement method, to conduct research and analysis.
3 Research Methods

Questionnaire design and distribution
Based on the review of relevant research results at home and abroad, the author proposed 4 hypotheses and 12 scale questions, from which a questionnaire was designed, through pre survey, the individual questions of the questionnaire were modified, and the model was slightly adjusted, finally forming three variables.The questionnaire consists of 13 measurement items derived from 3 variables.All questions in the questionnaire are scored by Likert 5, that is, according to the actual situation and real feelings of the respondents, they choose the appropriate corresponding degree of the measurement items from the five options of "fully agree", "agree", "general", "disagree" and "completely disagree", the measurement summary of all independent variables is shown in Table 1 [3][4].The questionnaire was designed through questionnaire stars, questionnaires were distributed through social media such as QQ space WeChat friends circle, WeChat group and microblog, a total of 324 questionnaires were collected over a period of 30 days, 314 valid ones were selected, with an effective rate of 96.91%, follow up analysis was conducted according to the questionnaire data.

Descriptive statistical analysis
1) Statistical analysis of basic personal information.314 valid questionnaires were collected in this study, of which 98 were male, accounting for 31.21% of the total, there were 216 female students, accounting for 68.79% of the total survey population.According to the proportion of men and women surveyed, there are more girls than boys.In terms of the distribution of English level, there are 18 students at junior middle school and below, accounting for 5.73%;There are 55 senior high school students with English proficiency, accounting for 17.51%;There are 163 people with CET-4, accounting for 51.91%;There are 53 students with CET-6, accounting for 16.88%; 12 people with professional level IV, accounting for 3.82%; There are 5 persons with Grade 8 professional level, accounting for 1.59%;There are 8 people with other English proficiency, accounting for 2.54% of the total sample.Due to the randomness of the questionnaire distribution, the distribution of the respondents is uneven, most of the respondents' English proficiency is CET-4, accounting for about half of the sample data, therefore, it can be seen that most of the respondents have bachelor's degree or above.Such groups have higher educational background, have enough time to spend freely, have less academic pressure and are more sensitive to the Internet [5].
2) Analysis of the use of English education and learning platform.Of the 314 sample data, 271 respondents have used the English education and learning platform, accounting for 86.3% of the total sample, it can be seen that most respondents have used the English education and learning platform.Of the 43 sample data that have not used the English education and learning platform, 14 respondents have not heard of online education, accounting for 32.6%;There are 11 people who think that English education and learning platform is not needed, accounting for 25.6%; Seven respondents disagreed with online education, accounting for 16.3%;There are 11 respondents who do not use the English education and learning platform for other reasons.It can be seen from this that most users who have not used the English education and learning platform have not heard of English education, have no demand for English online education, and a small number of survey objects do not recognize online education very much.
Among 271 respondents who have used the English education and learning platform, 178 (65.7%) of them use Kaochong English; The number of users of New Oriental is 104, accounting for 38.4%; 51Talk has 46 users, accounting for 17.0%;The number of users of Universal IELTS is 25, accounting for 9.2%; The number of people using American Union English is 28, accounting for 10.3%;The number of people using other English education and learning platforms is 16, accounting for 5.9% (the respondents have used multiple platforms, so the percentage is double counted).It can be seen from this ratio that the majority of people use Kaochong and New Oriental, as a well-known and old brand enterprise, the usage rate of New Oriental is lower than that of the newly created Kaochong English in recent years, it can be seen that the power of rising stars cannot be underestimated.
Analysis of continuous use of English online learning platform.
According to the user's experience in using the English education and learning platform, the author selected three questions in the form of a matrix scale to measure the continuous use of the platform.
The survey results show that the overall satisfaction of users who have used the English education and learning platform with the content quality of the platform is relatively high, the majority of users agree and fully agree with the platform, the satisfaction of users who have rich curriculum resources is the highest, the second is the quality of the course video, and the rationality of the relevant information and chapter arrangement of the course is the lowest.The survey results on the interactivity and comprehensive services of the platform show that users who have used the English online learning platform have the highest satisfaction with the platform system reminders, enabling users to fully agree.The interaction between teachers and students and between students is relatively low, but the overall satisfaction is OK.For the survey on the service of the platform, the satisfaction with free audition, reasonable price, perfect after-sales service, good evaluation system, layout design and platform compatibility is relatively balanced, moreover, users choose the most complete consent options.It can be seen that college students are highly satisfied with the service of the platform, which may strengthen users' willingness to continue using the platform [6][7].

Reliability Analysis and Validity Analysis
Reliability analysis.In this study, the reliability coefficient method will be used to measure the internal reliability of the question α.In a specific study, if the reliability index is greater than 0.7, it is generally assumed that the test item remains, and the test results are found to be consistent and stable.The results of the reliability tests are shown in Table 2. Table 2 shows that the Cronbach'a value of the difference is greater than 0.7 to 0.999, which shows that the research items of this study are stable, reliable and meet the evaluation needs, which is the next stage of analysis.can be sent.Reliability analysis for all variables was performed in the same way, and the results are presented in Table 3.It can be seen from Table 3 that the Cronbach'a value of each variable is greater than 0.7, so it can be explained that the stability of the four variables is strong, with high reliability, and it does not need to be revised, this shows that the credibility of this survey is very high, reaching the initial expectations, and the data obtained meet the needs of this study, which can be used for further analysis.
Validity analysis.Validity indicates that a measuring instrument can measure the value it is intended to measure.The more common the criteria, the more valid; Conversely, validity is low.KMO and Bartlett tests were performed because the author evaluates the validity of the data, i.e. each person is different and suitable for statistical analysis.Bartlett's test is used to measure the relationship between different items, and if the test is less than 0.01, the test equation can be rejected if the relationship between the equations is questionable, which indicates that there is a significant relationship between them.variables indicate testability.subject to change.KMO is a statistical test that compares simple correlation coefficients and partial correlation coefficients between variables.In general, the KMO statistic above 0.9 is the best result and suitable for statistical analysis, if it is less than 0.5, statistical analysis is not allowed.Analytical results for the validity questions of this study are presented in Table 4.The measured value of KM0 is 0.978, which is greater than 0.5, so this group of data is suitable for statistical analysis, as can be seen in Table 4.
Bartlett's test of sphericity has a significance probability of 0.000 and Sig <0.01, thus rejecting the null hypothesis that the correlation coefficient is equal to Eq., i.e., there is a significant relationship between the variables.Because the data for these questions are related and meet the research needs, the validity data is high and can be measured in the next step.
The author studied 3 factors, including 13 topics.Factor analysis is conducted between variables with the same factor, SPSS software will automatically classify the data according to the data concentration, and the data in the data set will be reclassified into one category to generate new variables, the rotation component matrix is obtained by factor analysis of the rotation matrix.Among them, some variable factors are not suitable to be attributed to the same factor, so the 13 variable factors can be split and combined again according to the data.Among them, the learning system in the platform's serviceability and perceived interactivity will remind in advance, the course supports free auditions, the price is reasonable, the after-sales service is perfect, the evaluation system is good, the layout design is reasonable, and the compatibility is strong, therefore, it is appropriate to attribute the same factor -platform serviceability; Rich curriculum resources, high quality curriculum videos, rich curriculum related information, and reasonable arrangement of curriculum chapters are displayed in the second column, which is the same as the original model, or is it a content quality factor; In the last column, the interaction between teachers and students and the effective interaction between students are classified as perceptual interaction factors, and a new model is obtained after reclassification [8].

Correlation and regression analysis
Correlation analysis.Pearson correlation coefficient and two-tailed test were used to examine the relationship between the selected factors.A significance level increased by theoretical analysis * indicates a significant relationship at the 0.05 (two-sided) level; * * plus indicates an interaction significant at the 0.01 level (two-sided).Table 5 is the correlation analysis between different variables based on the previous model.As shown in Table 5, p<0.01, and Pearson coefficients are greater than 0.5, therefore, the content quality, perceived interactivity, and platform serviceability are significantly related to the user's intention of continuous use at 0.01 level, and the preliminary hypothesis tests are all valid.
Regression analysis.After the correlation analysis, only the relationship and tightness of each variable were clarified, in order to understand the direction of the relationship between variables, regression analysis is required to further verify whether there is a causal relationship between variables.The author studies the causal relationship between each independent variable and intention to use, as well as between intention to use and the use behavior of English online learning platform users, and uses multiple regression analysis to verify the research hypothesis.In the regression model, the R value is 0.960, and the R square value is 0.921, which shows that the goodness of fit is good; In the F test, if the F value is larger, the regression effect is better, the results show that the F value is 1202.793and the probability of significance is 0.000, therefore, the regression effect is significant, as shown in Table 6.Table 6 shows that the probability of the coefficient of the test equation is 0, less than the significance level of 0.05, and 0 if the coefficient is different, the linear relationship between the explanatory variables and all the explanatory variables.is significant and a linear equation can be established.
Those.It can be concluded that the theoretical model has achieved certain reliability and validity, and the data and the model fit well.

Discussion
After the reliability analysis, validity analysis, correlation analysis and regression analysis of the data of the questionnaire, conclusions are drawn, and the conclusions are compared with the assumptions mentioned above, as shown in Table 7.

Relevant assumptions
Validation results H1: The early warning of the learning system has a significant impact on the user's willingness to continue using; H2: Effective interaction between teachers and students has a significant impact on users' willingness to use continuously; H3: Effective interaction between students has a significant impact on users' continuous use intention; H4: Rich curriculum resources have a significant impact on users' willingness to continue to use; H5: High quality of course video has a significant impact on users' continuous use intention; H6: Abundance of relevant information has a significant impact on users' continuous use intention; H7: Reasonable chapter arrangement has a significant impact on users' willingness to continue using; H8: Supporting free audition has a significant impact on users' continuous use intention; H9: Reasonable prices have a significant impact on users' willingness to continue using; H10: The improvement of after-sales service has a significant impact on users' willingness to continue using; H11: A good evaluation system has a significant impact on users' willingness to continue using; H12: Reasonable layout design has a significant impact on users' continuous use intention; According to the results of various hypothesis tests, this study analyzes these factors that affect the intention to use English online learning platforms continuously.First of all, perceived interactivity has a significant positive impact on users' willingness to use continuously, today, with the rapid development of Internet technology, there are various ways of communication between users, for example, online learning platforms such as WeChat, Weibo and QQ combine popular social software to achieve effective interaction and improve learning efficiency.Therefore, perceived interaction plays a significant role in promoting the willingness to use continuously.Secondly, content quality has a significant positive impact on users' continuous use [9][10].

Conclusion
Nowadays, the online learning platform is fiercely competitive, in order to attract more users, the online learning platform not only provides core courses for English learning, at the same time, it also provides other courses, such as oral English, translation, speeches and sharing meetings of famous people, the richness of such courses also promotes the continuous use of users.In addition, the rationality of video course recording and course arrangement also has an impact on users' continuous use.Therefore, content quality has a significant positive impact on users' willingness to continue using.Finally, platform serviceability has a significant positive impact on users' willingness to continue to use.The free trial courses provided by the platform enable users to have a general understanding of the courses, so as to help users make purchase decisions, the compatibility, layout design and after-sales service of the platform all affect users' satisfaction, the more considerate the services provided by the platform, the more effective it is in ensuring users' continuous use.

Figure 1 .
Figure 1.Factors Affecting the Continuous Use of Generalized Estimation Equation in College EnglishEducation Platform

Table 1 .
Summary of measurement problems

Table 2 .
Reliability Measurement Results

Table 3 .
Reliability coefficient of each variable

Table 4 .
Validity indicators of each variable

Table 5 .
Correlation Analysis of Continuous Use Intention and Various Factors

Table 6 .
Analysis of regression results of independent variables and intention to use continuously