Analysis method and device for causal relationship between variables, equipment and storage medium

Abstract

The invention provides an analysis method and device for a causal relationship between variables, equipment and a storage medium, and the method comprises the steps: obtaining an observation variable of a to-be-researched problem as an explicit variable, and carrying out the clustering of the observation variable through R-type clustering to generate a latent variable; setting a causal relationship between the explicit variable and the latent variable based on a theoretical basis; extracting latent variables with reason and result variables as intermediate variables; constructing a path diagram; and obtaining a load between the explicit variable and the latent variable and a path coefficient between the latent variables by adopting a partial least square method to finish path analysis. According to the method, an intermediary variable is extracted according to the causal relationship between the latent variables, a structural equation model path diagram is constructed, a partial least square method is introduced to construct a structural equation, and the relationship between the latent variables is analyzed more accurately.

Description

Technical Field

[0001] The present application belongs to the field of data statistical analysis, and in particular, relates to an analysis method, device, equipment and storage medium for causal relationships between variables. Background Art

[0002] In the field of data analysis and statistics, the study of latent variables has always been an important means to explore the internal mechanisms of complex systems. Latent variables cannot be directly observed, but they have a significant impact on observable variables (i.e., manifest variables) and are used to explain and predict the relationship between manifest variables. Traditionally, researchers rely on directly observed data to build models and analyze relationships, but this method often ignores those potential but important influencing factors, thereby limiting the explanatory power and predictive ability of the model.

[0003] In recent years, with the in-depth study of latent variables in psychology, sociology, economics and other disciplines, a series of methods aimed at revealing the relationship between latent variables have emerged. However, most of these methods face challenges such as difficulty in data collection, complex model construction and poor interpretability of results. Especially when dealing with data containing a large number of explicit variables and complex relationships between these explicit variables, how to effectively identify and extract latent variables, and further analyze the causal relationship between these latent variables, has become an urgent problem to be solved.

[0004] In the prior art, although some methods attempt to extract latent variables through factor analysis, principal component analysis, etc., these methods often focus on dimensionality reduction and feature extraction, while ignoring the causal relationship between latent variables. In addition, these methods rely on traditional multiple regression analysis methods when constructing structural equation models (SEM), which can lead to instability and inaccuracy of the model when dealing with small sample data or when there is multicollinearity. Summary of the invention

[0005] The purpose of the present application is to overcome the defects in the above-mentioned prior art and to provide a method, device, equipment and storage medium for analyzing the causal relationship between variables.

[0006] The present application provides a method for analyzing the causal relationship between variables, including:

[0007] Obtain observed variables according to the problem to be studied, and use the observed variables as explicit variables;

[0008] Clustering the manifest variables by R-type clustering to generate multiple data classes as latent variables;

[0009] Based on the theoretical basis of studying the problem to be studied, setting the explicit variables and the latent variables, and the causal relationship between the latent variables;

[0010] According to the causal relationship between the latent variables, a latent variable with a cause variable and a result variable is extracted as a mediating variable;

[0011] Constructing a path diagram according to the manifest variable and the latent variable, the causal relationship between the latent variables and the position of the mediating variable in the latent variable;

[0012] According to the path diagram, the partial least squares method is used to obtain the load between the manifest variable and the latent variable and the path coefficient between the latent variables, so as to complete the path analysis between the variables.

[0013] Optionally, clustering the explicit variables by R-type clustering to generate multiple data classes as latent variables includes:

[0014] The R-type cluster analysis algorithm was used to cluster the manifest variables;

[0015] According to the clustering results and the theoretical basis, the data class of each cluster is determined as a latent variable, and each latent variable is named to reflect the represented concept or feature.

[0016] Optionally, the path diagram includes: the causal relationship and direction between the manifest variables, the mediating variables, the cause variables and the result variables.

[0017] Optionally, the explicit variables are clustered by R-type clustering to generate multiple data classes, including:

[0018] The correlation coefficient matrix between the manifest variables is calculated, and the manifest variables are clustered into multiple data classes according to the numerical value of the correlation coefficient matrix, each of which represents a latent variable.

[0019] Optionally, constructing a path diagram according to the manifest variable and the latent variable, the causal relationship between the latent variables and the position of the mediating variable in the latent variable includes:

[0020] Graphical tools are used to draw path diagrams based on the causal relationship between explicit variables and latent variables, latent variables and the position of mediating variables.

[0021] Optionally, partial least squares methods are used to obtain loadings and path coefficients, including:

[0022] According to the path diagram, a set of equations including explicit variables, latent variables and mediating variables is constructed;

[0023] The partial least squares method is used to iteratively solve the equations to obtain the loadings between the manifest variables and the latent variables and the path coefficients between the latent variables.

[0024] Optionally, determining the score value of the latent variable according to the score value of the observed variable, the loading and the path coefficient includes:

[0025] Based on the loadings and path coefficients, the score equation of the latent variable was constructed;

[0026] Substitute the observed value of the manifest variable into the score equation to calculate the score of the latent variable.

[0027] The present application also provides a device for analyzing causal relationships between variables, including:

[0028] A construction module is used to obtain observed variables according to the problem to be studied, and use the observed variables as explicit variables;

[0029] A clustering module, used for clustering the explicit variables through R-type clustering to generate multiple data classes as latent variables;

[0030] A mediation module, used for extracting latent variables having cause variables and result variables as mediation variables according to the causal relationship between the latent variables;

[0031] A path module constructs a path diagram according to the manifest variable and the latent variable, the causal relationship between the latent variables and the position of the mediating variable in the latent variable;

[0032] The coefficient module is used to obtain the load between the explicit variable and the latent variable and the path coefficient between the latent variables according to the path diagram by using the partial least squares method to complete the path analysis between the variables.

[0033] The present application also provides a device for analyzing causal relationships between variables, including:

[0034] Memory;

[0035] A processor is used to call the computer executable program of the above-mentioned method for analyzing the causal relationship between variables stored in the memory, and execute: obtaining observed variables according to the problem to be studied, and taking the observed variables as explicit variables; clustering the explicit variables through R-type clustering to generate multiple data classes as latent variables; setting the explicit variables and the latent variables, and the causal relationship between the latent variables based on the theoretical basis for studying the problem to be studied; extracting the latent variables with cause variables and result variables as mediating variables according to the causal relationship between the latent variables; constructing a path diagram according to the explicit variables and the latent variables, the causal relationship between the latent variables and the latent variables, and the position of the mediating variable in the latent variables; according to the path diagram, using the partial least squares method to obtain the load between the explicit variables and the latent variables and the path coefficient between the latent variables, and complete the path analysis between the variables.

[0036] The present application also provides a storage medium, comprising: a computer executable program stored therein, wherein the computer executable program is used to be called by a processor to execute the steps of the above-mentioned method for analyzing the causal relationship between variables.

[0037] The beneficial effects of this application are:

[0038] The present application provides a method for analyzing the causal relationship between variables, including: obtaining observed variables according to the problem to be studied, and taking the observed variables as explicit variables; clustering the explicit variables through R-type clustering to generate multiple data classes as latent variables; setting the explicit variables and the latent variables, and the causal relationship between the latent variables and the latent variables based on the theoretical basis of studying the problem to be studied; extracting latent variables with cause variables and result variables as mediating variables according to the causal relationship between the latent variables; constructing a path diagram according to the explicit variables and the latent variables, the causal relationship between the latent variables and the latent variables, and the position of the mediating variable in the latent variables; according to the path diagram, using the partial least squares method to obtain the load between the explicit variables and the latent variables, the path coefficient between the latent variables, and complete the path analysis between the variables. The present application extracts mediating variables according to the causal relationship between the latent variables, and constructs a path diagram of the structural equation model based on the causal direction of the mediating variables, the cause variables and the result variables. Finally, the structural equation is constructed by introducing the partial least squares method, so as to achieve a more accurate analysis of the relationship between the latent variables. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 This is a schematic diagram of the critical care process based on intelligent analysis in this application;

[0040] Figure 2 This is a schematic diagram of the CRAM model application process in this application;

[0041] Figure 3 Schematic diagram of the intensive care system based on intelligent analysis in this application. DETAILED DESCRIPTION

[0042] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that various forms of implementing the present disclosure should not be limited by the embodiments set forth herein. On the contrary, the embodiments are provided to more thoroughly understand the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0043] Please refer to Figure 1 As shown, the present application provides a method for analyzing the causal relationship between variables, and the specific steps of the method include:

[0044] S101. Obtain observed variables according to the problem to be studied, and use the observed variables as explicit variables.

[0045] Structural equation modeling (SEM) is an advanced multivariate statistical analysis method that is used to simultaneously process the relationship between multiple independent and dependent variables and take into account the existence of measurement errors. SEM is suitable for dealing with complex causal relationships and latent variables (i.e., variables that cannot be directly observed), and reflects the changes in latent variables through multiple observed variables (i.e., variables that can be directly measured). For example: in the field of human resource management, the relationship between employee job satisfaction and organizational performance is analyzed; in the field of educational research, the impact of educational policies on students' academic performance is studied; in the field of social sciences, the relationship between social phenomena and mental health is studied, etc.

[0046] In order to study the path relationship between variables, we first set up a questionnaire based on the problem to be studied, for example: to study the relationship between variables in the job satisfaction of employees of an airline company.

[0047] A total of multiple questionnaire data were obtained through questionnaire surveys, and the questionnaires included 13 satisfaction survey-related questions, as shown in Table 1-1.

[0048] Table 1-1

[0049]

[0050] All questions are explicit variable questions, and are uniformly measured using a five-level scoring system. The corresponding meanings are shown in Table 1-2 below:

[0051] Table 1-2

[0052]

[0053]

[0054] Based on the above survey questions, a set of observed variables are obtained, which are the explicit variables.

[0055] S102: Clustering the explicit variables through R-type clustering to generate multiple data classes as latent variables.

[0056] R-type clustering is an analytical method for clustering explicit variables. The specific implementation steps include: calculating the correlation coefficient matrix between explicit variables, which reflects the degree of linear correlation between each explicit variable. According to the numerical value of the correlation coefficient matrix, an appropriate clustering algorithm (such as K-means, hierarchical clustering, etc.) is used to cluster the explicit variables to generate multiple data classes. Each data class represents a latent variable, which is a comprehensive reflection of the explicit variables to some extent. Finally, the clustering results should be explained and verified based on the theoretical basis to ensure the rationality and effectiveness of the clustering.

[0057] The naming of latent variables should reflect the concepts or characteristics they represent.

[0058] According to R-type cluster analysis (for explicit variables), for example, 13 explicit variables are clustered into 5 categories, namely: G1 = {x1, x2, x3, x4, x6, x 11 ,x 12 ,x 13},G2={x7}, G3={x8}, G4={x9,x 10}, G5 = {x5}. According to the meaning of measurable explicit variables, the five major types are: G1 company development, G2 work status, G3 communication channels, G4 learning and training, and G5 self-planning.

[0059] Furthermore, when studying the job satisfaction of the employees of the airline mentioned above, we can adopt the "hierarchy of needs theory" proposed by Maslow, that is, the five levels of human needs are ranked in order of priority (from low to high), namely physiological needs, safety needs, social needs, respect needs and self-actualization needs. When examining the employee satisfaction of the airline, we should also consider the hierarchy of needs from low to high.

[0060] Based on the results of R-type cluster analysis and the "demand hierarchy theory", this application sets three latent variables, whose demand levels from low to high are low-level demand: work environment ξ1, medium-level demand: personal ability and improvement ξ2 and high-level demand: company development prospects ξ3. At the same time, the research purpose is set as a latent variable describing overall satisfaction, overall satisfaction η1. As shown in Table 1-3:

[0061] Table 1-3

[0062]

[0063]

[0064] S103. Based on the theoretical basis for studying the problem to be studied, setting the explicit variables and the latent variables, and the causal relationship between the latent variables.

[0065] The structural equation model includes: a measurement equation and a structural equation. The measurement equation is used to express the relationship between explicit variables and latent variables, and the structural equation is used to express the relationship between latent variables.

[0066] This application infers the causal process of the structural model based on multiple regression analysis and correlation analysis.

[0067] According to the theoretical basis of the research problem, such as the "hierarchy of needs theory", the causal relationship of each latent variable is set.

[0068] Taking company development prospects, personal ability and improvement, and working environment as independent variables and overall satisfaction as dependent variable, a multiple regression model was established using SPSS software. The model expression is:

[0069] Employee satisfaction = 0.462 + 0.321 × company development prospects + 0.138 × personal ability and improvement + 0.409 × work environment.

[0070] Assume that the company's development prospects, personal ability and improvement, and working environment are the cause, and overall satisfaction is the result. Therefore, the company's development prospects, personal ability and improvement, and working environment are exogenous latent variables, and overall satisfaction is an endogenous latent variable.

[0071] Among them, based on multiple regression analysis, the correlation coefficient table between latent variables is shown in Table 3-5:

[0072] Table 3-5 Correlation coefficients of latent variables

[0073]

[0074] As shown in the above table, the correlation coefficients between the exogenous latent variables are mostly around 0.5, so the relationship between the exogenous latent variables is an association relationship.

[0075] like Figure 2 As shown, a first structural equation path diagram is constructed based on the association relationship.

[0076] In the present application, the measurement equation includes a reflective measurement model and a formative measurement model. In the reflective measurement model, the one-way arrow representing the causal relationship points from the latent variable to the measurable variable; in the formative measurement model, the one-way arrow representing the causal relationship points from the measurable variable to the latent variable. The difference between the reflective measurement model and the formative measurement model lies in the judgment of the causal relationship between the latent variable and the measured variable. If the transformation of the latent variable causes the change of the manifest variable, it means that the latent variable is the cause and the manifest variable is the effect, and a reactive measurement model is constructed; if the change of the manifest variable causes the change of the latent variable, it means that the manifest variable is the cause and the latent variable is the effect, and a formative measurement model is constructed.

[0077] Furthermore, based on the "hierarchy of needs theory" of satisfaction and the first structural equation path diagram mentioned above, a hierarchical structure with low-level satisfaction as the cause and medium-level and high-level satisfaction as the result is set. The assumptions of the correlation between the three endogenous variables originally set are changed to: "work environment → personal ability and improvement", "work environment → company development prospects" and "personal ability and improvement → company development prospects".

[0078] So far, the present application has inferred the causal relationship between the explicit variables and the latent variables, and obtained the causal relationship between the explicit variables and the latent variables, and between the latent variables.

[0079] S104. According to the causal relationship between the latent variables, extract the latent variables having the cause variables and the result variables as the mediating variables.

[0080] The causal variable, also known as the independent variable or antecedent variable, is the variable that affects other variables. In a causal relationship, it is considered the "cause", that is, the variable that causes other variables to change. The outcome variable, also known as the dependent variable or consequence variable, is the variable that is affected by other variables. In a causal relationship, it is considered the "effect", that is, the variable caused by the causal variable.

[0081] A mediating variable is a variable between a causal variable and an outcome variable. It is located between the causal variable and the outcome variable and transmits or mediates the effect of the causal variable on the outcome variable. In other words, the causal variable indirectly affects the outcome variable through the mediating variable.

[0082] In this application, based on the relationship between the latent variables, "personal ability and improvement" is the mediating variable in the relationship between "work environment" and "company development prospects"; "personal ability and improvement" is the mediating variable in the relationship between "work environment" and "overall satisfaction"; "company development prospects" is the mediating variable in the relationship between "personal ability and improvement" and "overall satisfaction".

[0083] S105 , constructing a path diagram according to the manifest variables and the latent variables, the causal relationship between the latent variables, and the position of the mediating variable in the latent variables.

[0084] In this application, the path diagram is used to show the causal relationship and direction of explicit variables and latent variables. Construction details include: using graphical tools (such as Visio, Graphviz, etc.) to draw the path diagram. In the path diagram, the node represents the variable (explicit variable or latent variable), and the arrow represents the direction of the causal relationship. Draw a complete path diagram based on the causal relationship between explicit variables and latent variables, latent variables and latent variables, and the position of the mediating variable.

[0085] like Figure 3 As shown in Figure 1, based on the causal relationship between the manifest variable and the latent variable, between the latent variable and the latent variable, and the determined mediating variable, a path diagram containing the mediating variable is constructed. Among them, "personal ability and improvement" and "company development prospects" are mediating variables, and the arrows indicate the direction of the causal relationship.

[0086] Based on this, two optimization paths can be obtained: work environment → personal ability and improvement → company development prospects → overall satisfaction; work environment → company development prospects → overall satisfaction.

[0087] S106. According to the path diagram, partial least squares method is used to obtain the load between the manifest variable and the latent variable and the path coefficient between the latent variables, so as to complete the path analysis between the variables.

[0088] Partial least squares is an iterative solution method used to obtain the loadings between explicit and latent variables and the path coefficients between latent variables. A system of equations containing explicit variables, latent variables, and mediating variables is constructed based on the path diagram. The system of equations is iteratively solved using partial least squares to obtain estimates of the loadings and path coefficients. During the iteration process, weights and path coefficients are estimated by minimizing the prediction error. The final loadings and path coefficients should be interpreted and verified based on the theoretical basis to ensure their accuracy and validity.

[0089] Specifically, the calculation process of partial least squares method is:

[0090] From the standardized independent variable (X) and dependent variable (Y), the first pair of components (called components t and u) are first extracted. Using the Lagrange multiplier method, the expressions of components t and u are solved while satisfying the goals of maximum variance and maximum correlation.

[0091] Using the least squares method, a regression model is built for components t and u. This model describes how component t predicts component u. Substitute the relationship between component t and the original independent variable X back into the regression model of t and u to obtain the regression model between the dependent variable Y and the independent variable X.

[0092] After the first iteration, the residual matrix is ​​used to replace the original data, and the above steps of component extraction, target solution, regression modeling and back substitution are repeated until r components are extracted, where r is the rank of the independent variable X matrix.

[0093] Based on the partial least squares method, a structural equation model (PLS-SEM) is constructed:

[0094] In the structural equation model, what needs to be obtained is the relationship between latent variables (or structural variables), and the relationship between these latent variables and observed indicators (or measured variables). The application of PLS ​​in SEM is mainly reflected in the following aspects:

[0095] Measurement Model:

[0096] In the measurement model, PLS is used to estimate the relationship between the observed indicators and their corresponding latent variables. This is achieved by calculating loadings (or weights), which reflect the contribution of the observed indicators to the latent variables.

[0097] PLS handles the problem of multicollinearity between observed indicators, thereby estimating the loadings more accurately.

[0098] Structural model:

[0099] In the structural model, this application uses PLS to estimate the relationship between latent variables. These relationships are expressed as path coefficients or path weights, which reflect the degree of mutual influence between latent variables.

[0100] Through the components extracted by PLS, the present application can establish a regression model between latent variables to estimate the path coefficients.

[0101] The specific steps are as follows:

[0102] Step 1: Standardize all raw data to ensure that the data is on the same scale. Initialize all weight vectors. In the initial stage of the algorithm, an initial weight is assigned to each explicit variable of the structural latent variable. The initial weight is generated randomly or set based on prior knowledge or assumptions. The initial weight will serve as the starting point for subsequent iterative calculations.

[0103] Step 2: Parameter estimation.

[0104] Structural latent variables are not directly observable, but they have an impact on other variables in the model. In order to estimate the model, scores need to be generated for these structural latent variables. The scores are calculated based on the initial weight vector and the values ​​of the observable variables (also called indicators or manifest variables). The scores of structural latent variables are calculated by taking a weighted sum (or weighted average) of the initial weight vector and the values ​​of the corresponding observable variables.

[0105] Estimate internal structural model weights:

[0106] Centroid: The centroid (positive or negative) is determined based on the correlation coefficients between the structural latent variables. The centroid is a position or direction indicator of the latent variable in the relationship network. It is determined based on the correlation coefficients between the structural latent variables. If the correlation coefficient between two structural latent variables is positive, then they have similar trends and can therefore be considered to be in the same "direction". Conversely, if the correlation coefficient is negative, then they have opposite trends.

[0107] Factorial: Analyze the relationship between two structures. Factorial analysis involves exploring the relationship between two structures. It includes analyzing the direct effect, indirect effect, and total effect between them. Factorial analysis helps to understand how the structural latent variables influence each other and how this influence is propagated in the model.

[0108] Path weight: The weight of the relationship between two constructs is calculated using a partial least squares regression model. The path weight represents the strength and direction of the relationship between the latent variables of the constructs. In SEM, the weight is calculated using partial least squares regression (PLS). PLS is an iterative method that estimates the weight by minimizing the prediction error at each step.

[0109] Reestimate Path Weights: Update path weights using the newly calculated structural weights.

[0110] Estimate measurement model weight loadings:

[0111] Mode A (reflective measurement model): The standardized correlation between the construct and each indicator is calculated.

[0112] Mode B (Formative Measurement Model): Partial least squares regression model coefficients were used as the relationship between the construct and each indicator.

[0113] Check convergence: If the sum of the changes in the measured model weights is less than the stopping criterion, the iteration stops.

[0114] Step 3: Final model.

[0115] The weight vector obtained in the previous step is used to calculate the final score of the structural latent variable. The final score represents the actual value or estimated value of the structural latent variable and can be used for further analysis and interpretation. Specifically, the calculation process multiplies the values ​​of the observed variables by their respective weights and then adds these weighted values. This process reflects the contribution of the observed variables to the structural latent variable.

[0116] The path coefficients of the internal structural model were calculated using partial least squares:

[0117] In the partial least squares-structural equation model (PLS-SEM), the path coefficients of the internal structural model represent the strength and direction of the relationship between the structural latent variables.

[0118] The calculation of path coefficients is done through an iterative process of partial least squares. In each iteration, a new set of components that maximize the covariance between the structural latent variables is extracted and used to estimate the path coefficients.

[0119] Finally, through multiple iterations, stable path coefficient estimates are obtained, which reflect the actual relationship between the structural latent variables.

[0120] The measurement model loadings were used as correlation coefficients between the constructs and each indicator to interpret the model:

[0121] The measurement model loadings represent the degree of association between the observed variables (or indicators) and their corresponding structural latent variables. In PLS-SEM, these loadings are estimated by weight vectors in an iterative process.

[0122] Taking the measurement model loading as the correlation coefficient between the structure and each indicator can help this application understand how the observed variables affect the structural latent variables and the specific manifestations of these influences in the model.

[0123] These correlation coefficients can also be used to interpret the results of the model. For example, by comparing the load sizes of different observed variables on the same structural latent variable, it can be determined which observed variables have a more significant impact on the structural latent variable.

[0124] The present application also provides a device for analyzing causal relationships between variables, including:

[0125] A construction module is used to obtain observed variables according to the problem to be studied, and use the observed variables as explicit variables;

[0126] A clustering module, used for clustering the explicit variables through R-type clustering to generate multiple data classes as latent variables;

[0127] A mediation module, used for extracting latent variables having cause variables and result variables as mediation variables according to the causal relationship between the latent variables;

[0128] A path module constructs a path diagram according to the manifest variable and the latent variable, the causal relationship between the latent variables and the position of the mediating variable in the latent variable;

[0129] The coefficient module is used to obtain the load between the explicit variable and the latent variable and the path coefficient between the latent variables according to the path diagram by using the partial least squares method to complete the path analysis between the variables.

[0130] The present application also provides a device for analyzing causal relationships between variables, including:

[0131] Memory;

[0132] A processor is used to call the computer executable program of the above-mentioned method for analyzing the causal relationship between variables stored in the memory, and execute: obtaining observed variables according to the problem to be studied, and taking the observed variables as explicit variables; clustering the explicit variables through R-type clustering to generate multiple data classes as latent variables; setting the explicit variables and the latent variables, and the causal relationship between the latent variables based on the theoretical basis for studying the problem to be studied; extracting the latent variables with cause variables and result variables as mediating variables according to the causal relationship between the latent variables; constructing a path diagram according to the explicit variables and the latent variables, the causal relationship between the latent variables and the latent variables, and the position of the mediating variable in the latent variables; according to the path diagram, using the partial least squares method to obtain the load between the explicit variables and the latent variables and the path coefficient between the latent variables, and complete the path analysis between the variables.

[0133] The present application also provides a storage medium, comprising: a computer executable program stored therein, wherein the computer executable program is used to be called by a processor to execute the steps of the above-mentioned method for analyzing the causal relationship between variables.

[0134] Although the present invention is disclosed as above by the embodiments, it is not intended to limit the protection scope of the present invention. Any changes and modifications made by any technician familiar with the technology without departing from the concept and scope of the present invention should fall within the protection scope of the present invention.

Claims

1. A method for analyzing causal relationships between variables, characterized in that: include: Obtain observed variables according to the problem to be studied, and use the observed variables as explicit variables; Clustering the manifest variables by R-type clustering to generate multiple data classes as latent variables; Based on the theoretical basis of studying the problem to be studied, setting the explicit variables and the latent variables, and the causal relationship between the latent variables; According to the causal relationship between the latent variables, a latent variable with a cause variable and a result variable is extracted as a mediating variable; Constructing a path diagram according to the manifest variable and the latent variable, the causal relationship between the latent variables and the position of the mediating variable in the latent variable; According to the path diagram, the partial least squares method is used to obtain the load between the manifest variable and the latent variable and the path coefficient between the latent variables, so as to complete the path analysis between the variables.

2. The method for analyzing the causal relationship between variables according to claim 1, characterized in that: The manifest variables are clustered by R-type clustering to generate multiple data classes as latent variables, including: The R-type cluster analysis algorithm was used to cluster the manifest variables; According to the clustering results and the theoretical basis, the data class of each cluster is determined as a latent variable, and each latent variable is named to reflect the represented concept or feature.

3. The method for analyzing the causal relationship between variables according to claim 1, characterized in that: The path diagram includes: the causal relationship and direction between the manifest variables, the mediating variables, the cause variables and the result variables.

4. The method for analyzing the causal relationship between variables according to claim 1, characterized in that: The explicit variables are clustered by R-type clustering to generate multiple data classes, including: The correlation coefficient matrix between the manifest variables is calculated, and the manifest variables are clustered into multiple data classes according to the numerical value of the correlation coefficient matrix, each of which represents a latent variable.

5. The method for analyzing the causal relationship between variables according to claim 1, characterized in that: According to the manifest variable and the latent variable, the causal relationship between the latent variables and the position of the mediating variable in the latent variable, a path diagram is constructed, including: Graphical tools are used to draw path diagrams based on the causal relationship between explicit variables and latent variables, latent variables and the position of mediating variables.

6. The method for analyzing causal relationships between variables according to claim 1, characterized in that: The partial least squares method was used to obtain the loadings and path coefficients, including: According to the path diagram, a set of equations including explicit variables, latent variables and mediating variables is constructed; The partial least squares method is used to iteratively solve the equations to obtain the loadings between the manifest variables and the latent variables and the path coefficients between the latent variables.

7. The method for analyzing causal relationships between variables according to claim 1, characterized in that: Determining the score value of the latent variable according to the score value of the observed variable, the loading and the path coefficient includes: Based on the loadings and path coefficients, the score equation of the latent variable was constructed; Substitute the observed value of the manifest variable into the score equation to calculate the score of the latent variable.

8. An analysis device for causal relationships between variables, characterized in that: include: A construction module is used to obtain observed variables according to the problem to be studied, and use the observed variables as explicit variables; A clustering module, used for clustering the explicit variables through R-type clustering to generate multiple data classes as latent variables; A mediation module, used for extracting latent variables having cause variables and result variables as mediation variables according to the causal relationship between the latent variables; A path module constructs a path diagram according to the manifest variable and the latent variable, the causal relationship between the latent variables and the position of the mediating variable in the latent variable; The coefficient module is used to obtain the load between the explicit variable and the latent variable and the path coefficient between the latent variables according to the path diagram by using the partial least squares method to complete the path analysis between the variables.

9. An analytical device for causal relationships between variables, characterized in that: include: Memory; A processor, for calling a computer executable program of a method for analyzing a causal relationship between variables as described in any one of claims 1 to 7 stored in the memory, and executing: obtaining observed variables according to the problem to be studied, and taking the observed variables as explicit variables; clustering the explicit variables through R-type clustering to generate multiple data classes as latent variables; setting the explicit variables and the latent variables, and the causal relationship between the latent variables based on the theoretical basis for studying the problem to be studied; extracting latent variables with cause variables and result variables as mediating variables according to the causal relationship between the latent variables; constructing a path diagram according to the explicit variables and the latent variables, the causal relationship between the latent variables and the latent variables, and the position of the mediating variable in the latent variables; according to the path diagram, using the partial least squares method to obtain the load between the explicit variables and the latent variables and the path coefficient between the latent variables, and completing the path analysis between the variables.

10. A storage medium, characterized in that: include: A computer executable program is stored, and the computer executable program is used to be called by a processor to execute the steps of the method for analyzing the causal relationship between variables as described in any one of claims 1 to 7.

Images