A regression analysis study on the cognitive change pattern of digital art display on college students’ thought leadership process

1)

Interactivity

The interactivity shown by digital media art mainly lies in the fact that people can participate in it. In the traditional display design, the display design is the art of a few people, and the exhibitors can only passively participate in it to enjoy the art created by the creators, and the distance between art and life is far. But after the application of digital media art in display design, the exhibitors can realize the communication and interaction with the three-dimensional picture through digital media technology, so that the whole process of display from silent to sound, from single media to multimedia, from the exhibition of real objects to virtual mirror show, from passive participation to active interaction, shaping the form of art exhibitions in our country in the form of a new form.

Under the digital era, college students’ lives are full of all kinds of digital information, and they have long been accustomed to receiving and acquiring information through social media, virtual reality, and other digital platforms. The characteristics of digital media with various forms of art and strong interactivity are quite consistent with the cognitive habits of college students, so it can guide them to form a correct and comprehensive cognition of civic politics in civic politics activities. 1)

Improve the depth of ideological and political cognition

The traditional form of civic education activities is based on classroom lectures, with the speaker explaining the content of civic politics alone. The form of communication is too monolithic and programmed, and it is traditional indoctrination education, which is difficult to stimulate students’ interest and enthusiasm in learning civic politics. Digital media art integrates visual design, animation, audio production, interactive technology and other elements, through intuitive images, animation and other means to diversify the display of Civic Politics content, in the auditory and visual can better attract the attention of students. Through artistic interpretation of socialist core values, revolutionary history, and other ideological themes, abstract ideological concepts are visualized and made more contagious, helping students to resonate emotionally. In addition, in digital media, students use virtual reality technology to interact and communicate with works of art related to ideology and politics. Breaking the subject limitations, improving students’ participation in civic politics activities, and making their understanding of civic politics theory more comprehensive. Not only that, digital media in civic and political display content is also timely. It combines the concept of ideological knowledge with current events in society, which prompts college students to think about social and political phenomena and then understand the meaning of ideological theory. Thus, it guides students to think and understand the content of ideological and political theories at a deeper level, and deepens their understanding of ideological and political theories.

The univariate linear regression modeling method studies the linear relationship between a single dependent variable and a single independent variable, which is suitable for relatively simple series prediction because of its fewer influencing factors and considerations. However, in real life, many things are closely related to the surrounding factors, for example, the development of the power industry involves not only the coal industry, but also transportation, environmental protection, industrial output, GDP and economic layout, and a variety of factors work together on the target. At this point, the univariate linear regression model is no longer used, and multiple linear regression is needed to analyze and study. Multiple linear regression reflects the linear relationship between multiple variables, which has a wider scope of application, and the results of linear prediction are closer to the actual values.

Multiple linear regression method assumes that the research objectives are influenced by multiple factors x₁, x₂, …x_m, and if the dependent variable y is linearly related to multiple other independent variables and there is a certain linear relationship, the multiple linear relationship between dependent variables y and x₁, x₂, …x_m is shown in equation (1): (1)yi=β1xi1+β2xi2+…….+βmxim+εi$${y_i} = {\beta _1}{x_{i1}} + {\beta _2}{x_{i2}} + \ldots \ldots . + {\beta _m}{x_{im}} + {\varepsilon _i}$$

If we take the observation of x_i1 to be constant 1, and x_n1 = 1 for any i, we obtain equation (2): (2)yi=β1+β2xi2+…….+βmxim+εi$${y_i} = {\beta _1} + {\beta _2}{x_{i2}} + \ldots \ldots . + {\beta _m}{x_{im}} + {\varepsilon _i}$$

Its matrix form is equation (3) (3)Y=Xβ+ε$$Y = X\beta + \varepsilon$$

To estimate parameter β, we use the least square method and set the residual difference between the observations and the model estimates to be E, then E=Y−Y^$$E = Y - \widehat Y$$, where Y^=Xβ$$\widehat Y = X\beta$$. The estimate of the vector of regression coefficients β, collapsed according to the least square method, is given in equation (4): (4)β^=(XTX)−1XTY$$\hat \beta = {\left( {{X^T}X} \right)^{ - 1}}{X^T}Y$$

The multiple linear regression model, like the homogeneous linear regression model, needs to test whether there is a corresponding linear relationship between multiple variables and the research target. If the significance test of multiple linear is not obvious, it means that there is no obvious linear relationship between the research target and other factors, and it is impossible to use multiple linear regression for prediction. So it is said that significance is the premise and foundation of using multiple linear for prediction. There are three common tests for multiple linear regression models: R test, t test, F test and DW test. 1)

R-test. The R test utilizes the complex correlation coefficient to test the linear relationship between each of the independent variables x₁, x₂, …x_m and the dependent variable y in the target series of the study. The R test is essentially similar to the test in the univariate linear regression model, and can be obtained by analyzing the total deviation and obtaining the formula (5) for the R² of the multiple linear regression model: (5)R=1−∑(yi−y^i)2∑(yi−y¯)2$$R = \sqrt {1 - \frac{{\sum {{{\left( {{y_i} - {{\hat y}_i}} \right)}^2}} }}{{\sum {{{\left( {{y_i} - \bar y} \right)}^2}} }}}$$

It can be shown that the F statistic obeys a F distribution with (m−1,n−m)$$\left( {m - 1,n - m} \right)$$ degrees of freedom, and in conjunction with the level of significance a given in the research objectives, a check of the F distribution table yields the critical value Fa(m−1,n−m)$${F_a}\left( {m - 1,n - m} \right)$$ If F>Fa(m−1,n−m)$$F > {F_a}\left( {m - 1,n - m} \right)$$, the hypothesis H₀ is rejected, and the regression between a set of independent variables x₁, x₂, …x_m and the dependent variable y is considered to be significant, and conversely, not significant. 3)

t test. The t test is a method of testing whether the hypotheses H₀ : β_j = 0, j = 1, 2, …m are valid by t statistics for each coefficient of the desired regression model one by one. The steps of the T test are as follows: first the standard error of the assessment is calculated as in equation (7): (7)S=∑(yi−y^i)n−m$$S = \sqrt {\frac{{\sum {\left( {{y_i} - {{\hat y}_i}} \right)} }}{{n - m}}}$$

Next the sample standard deviation is to be calculated from the imputation: SB^j=Cii*S$${S_{{{\hat B}_j}}} = \sqrt {{C_{ii}}} *S$$ and the t statistic is calculated. Finally, hypothesis H₀ : β_j = 0 is established. If |tj|≥ta/2(n−m)$$\left| {{t_j}} \right| \geq {t_{a/2}}(n - m)$$ is established, hypothesis H₀ is rejected, indicating that x_j has a significant effect on y; conversely, the hypothesis is established and H₀ : β_j = 0 accepted, indicating that x_j has no relationship for y or the relationship is so small that it can be ignored, so the factor can be disregarded. 4)

DW test: Durbin Watson criterion is a common test of multiple linear correlation, which uses the regression model and residuals to calculate the DW statistic, and the formula for DW is equation (8): (8)DW=∑i=2n(ei−ei−1)2∑i=1nei2$$DW = \frac{{\sum\limits_{i = 2}^n {{{\left( {{e_i} - {e_{i - 1}}} \right)}^2}} }}{{\sum\limits_{i = 1}^n {e_i^2} }}$$

Through equation (6) we can see that the DW value is between 0 and 4. According to the DW statistic, the model is tested for autocorrelation as follows:

First use the least square method to find the regression model and residuals e_i and calculate the DW statistic;

Secondly, establish hypothesis H₀ : ρ₁ = 0, i.e., assume that there is no autocorrelation in the regression model;

Finally, according to the given test level and the number of independent variables m from the DW test table to find the corresponding critical values d_L, d_U, and the use of Table 1 to identify the test conclusions:

The steps for calculating the prediction intervals for the multiple linear regression model are as follows: 1)

Calculate the standard error of the estimate: S=∑(yi−y^i)2n−m$$S = \sqrt {\frac{{\sum {{{\left( {{y_i} - {{\hat y}_i}} \right)}^2}} }}{{n - m}}}$$

Here X₀ is a vector of influence factor data, and in practical forecasting, S is generally used instead of S₀ to approximate the prediction intervals.

The initial test subjects were 50 in total, namely, 10 undergraduates (with any major), 20 graduate students majoring in Marxist theory, 10 graduate students not majoring in Marxist theory, and 10 professors of Civics and Political Science courses. First, the preliminary designed questionnaire was sent to these 50 testers, so that they could review the questionnaire with their respective cognitive levels and professional perspectives, and mark the unreasonable questions as well as the possible problems in the overall design of the questionnaire. Subsequently, according to the results of the questionnaire, targeted interviews were conducted focusing on 10 undergraduate students. Through these interviews, problems with the questions and the questionnaire as a whole were clarified. Finally, the opinions from the interviews were integrated and the experts specialized in Civics and Political Science were consulted, according to the experts’ opinions, the unclear, ambiguous and difficult-to-understand topics were deleted or amended, and finally 60 topics were formed and established.

To ensure a rich, hierarchical, and balanced sample, a variety of sampling methods were employed, and 1500 undergraduate students enrolled in a university were selected for testing. Table 2 shows the basic composition of the sample. The research sample totaled 1500 undergraduate students at school, and the test subjects were students who had participated in digital media art forms of ideological activities. 1500 questionnaires were distributed and 1469 were returned, with a 97.93% return rate for the questionnaires. According to the screening criteria for omission and nonsense answers, 103 invalid questionnaires were eliminated, and 1423 valid questionnaires were received, with an effective rate of 94.87%.

Based on the literature study, “the leading role of digital media art on the level of college students’ ideology and politics” needs empirical research to verify its reliability and relevance. In the compilation of the questionnaire, the richness of the content, the diversity of the presentation forms and the participation of students in digital media art were taken as important predictions, and a subscale of the influencing factors of digital media art on the ideological and political level of college students was compiled to verify the reliability and appropriateness of the “leading role and cognitive change mode” in theoretical research, and to test the influence of the three factors on the ideological and political cognitive level of college students.

In order to understand the predictive power of content richness, diversification of presentation forms, and students’ high degree of participation on the total scale of students’ Civic and Political Awareness Level and the accessibility of each dimension subscale, the predictive power of these variables was tested by using the stepwise multiple regression method, and the independent variables were selected into the regression model according to the statistical criterion in order to find out the most predictive power for the validity variables among the three independent variables of content richness, diversification of presentation forms, and students’ high degree of participation to construct an optimal regression analysis model. The independent variable with the most predictive power for the criterion variables is used to construct an optimal regression analysis model.