Causal relation analysis method and causal relation analysis program

Abstract

To provide a causal graph that makes it easier to explain a relation between an objective variable and explanatory variables.SOLUTION: A causal relation analysis includes: accepting input of an objective variable and explanatory variables; executing a causal search using the objective variable and the explanatory variables; identifying causal relations between respective variables; estimating a delay in the timing of occurrence of causality in each of the identified causal relations; and creating and outputting a causal graph that represents the identified causal relations and the estimated delay.SELECTED DRAWING: Figure 14

Description

Technical Field

[0001] The present invention relates to a causal relationship analysis method and a causal relationship analysis program.

Background Art

[0002] In corporate management, the importance of improving corporate value through sustainable management (management that aims to improve the sustainability of a business by considering the sustainability of the environment, society, and corporate governance) is increasing.

[0003] In order to improve corporate value, it is important to analyze how the effects that occur when capital is invested in basic items such as finance and non-finance spread to non-financial information such as materiality (key issues), ESG indicators, and corporate value. This analysis result can be utilized for management decision-making and can also be used as evidence for various stakeholders and analysts.

[0004] Conventionally, there is a known technique for extracting from the timing of capital investment in an explanatory variable (for example, the ratio of female executive employees, etc.) when an effect appears (spreads) to an objective variable (for example, PBR: price-to-book ratio) after several years (see, for example, Patent Document 1, etc.).

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0006] Patent Document 1 discloses a technique for performing a correlation analysis between corporate value (such as PBR, etc.) and various ESG indicators while shifting the years, and extracting from the timing of ESG capital investment when an effect appears on corporate value after several years.

[0007] Generally, there is a mediating variable between the target variable and the explanatory variable, and it is considered that there is an indirect causal relationship between the target variable and the explanatory variable through the mediating variable. However, in the above Patent Document 1, since a direct correlation analysis between the target variable and the explanatory variable is performed, it is difficult to explain the validity and reasons of the causal relationship from the explanatory variable to the target variable.

[0008] In one aspect, an object of the present invention is to provide a causal relationship analysis method and a causal relationship analysis program capable of providing a causal graph that facilitates the explanation of the relationship between a target variable and an explanatory variable.

Means for Solving the Problems

[0009] In one embodiment, a causal relationship analysis method is a causal relationship analysis method in which a computer executes a process of receiving an input of a target variable and an explanatory variable, performing a causal search using the target variable and the explanatory variable, specifying a causal relationship between each variable, estimating a delay in the occurrence timing of causality in each of the specified causal relationships, creating a causal graph expressing the specified causal relationship and the estimated delay, and outputting the causal graph.

Effects of the Invention

[0010] It is possible to provide a causal graph that facilitates the explanation of the relationship between a target variable and an explanatory variable.

Brief Description of the Drawings

[0011]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

DETAILED DESCRIPTION OF THE INVENTION

[0012] Hereinafter, an embodiment of the analyzer will be described in detail with reference to FIGS. 1 to 14.

[0013] The analysis device 10 (Fig. 1) of the present embodiment is a device that creates a causal graph showing the causal relationship between the objective variable and the explanatory variable determined by the analyst and presents it to the analyst. More specifically, the analysis device 10 receives the input of the objective variable determined by the analyst, the explanatory variable (first explanatory variable) known to have a causal relationship with the objective variable, and the second explanatory variable (latent explanatory variable) whose contribution to the objective variable is unknown by human judgment. Then, the analysis device 10 performs causal search between the received variables and creates a causal graph between the first explanatory variable and the objective variable. In creating this causal graph, the analysis device 10 also considers the effect manifestation delay value (described later) between variables. Hereinafter, the analysis device 10 will be described in detail.

[0014] Fig. 1 shows the hardware configuration of the analysis device 10. The analysis device 10 is an information processing device such as a PC (Personal Computer). As shown in Fig. 1, the analysis device 10 includes a CPU (Central Processing Unit) 190, a ROM (Read Only Memory) 192, a RAM (Random Access Memory) 194, a storage (e.g., SSD (Solid State Drive) or HDD (Hard Disk Drive), etc.) 196, a network interface 197, a display unit 193, an input unit 195, and a drive 199 for a portable storage medium, etc. Each component of the analysis device 10 is connected to a bus 198. In the analysis device 10, the functions of each part shown in Fig. 2 are realized by the CPU 190 executing a program stored in the ROM 192 or the storage 196 (including a causal relationship analysis program), or a program read by the drive 199 for a portable storage medium from the portable storage medium 191. Note that the functions in Fig. 2 may also be realized by an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

[0015] Figure 2 shows a functional block diagram of the analyzer 10. As shown in Figure 2, in the analyzer 10, the CPU 190 functions as an input reception unit 12, a causal relationship exploration unit 14, an effect manifestation delay estimation unit 16, an integration processing unit 18, and an output unit 20 by executing a program.

[0016] The input reception unit 12 receives the input of the target variable, the first explanatory variable, and a plurality of second explanatory variables (latent explanatory variables). The analyst inputs the explanatory variable known to affect the target variable as the first explanatory variable, and inputs the variable whose influence on the target variable is unknown but likely to be related as the second explanatory variable. Figure 3 shows an example of the data structure of the information (input information) input by the analyst and received by the input reception unit 12. In the example of Figure 3, it is assumed that in the input information, eight first explanatory variables (x1 to x8) are set for the target variable y, and a predetermined number of second explanatory variables (v1, v2, …) are set.

[0017] The causal relationship exploration unit 14 calculates the presence or absence and direction of the causal relationship between variables based on the data of the target variable, the first explanatory variable, and the second explanatory variable. The causal relationship exploration unit 14 performs causal relationship exploration by, for example, the methods described in "A Linear Non-Gaussian Acyclic Model for Causal Discovery" Shohei Shimizu, et al. Journal of Machine Learning Research 7 (2006) 2003-2030 URL:https: / / www.cs.helsinki.fi / group / neuroinf / lingam / JMLR06.pdf and "Causal Inference by Structural Equation Modeling: Recent Developments in Causal Structure Search" (The Japanese Society for Psychometrics 2012) URL:https: / / www.slideshare.net / sshimizu2006 / bsj2012-tutorial-finalweb.

[0018] FIG. 4(a) shows an example of the causal exploration result by the causal exploration unit 14. As a result of the causal exploration, a causal coefficient (an index representing the strength of the causal relationship) between the target variable and each explanatory variable, and a causal coefficient between the explanatory variables are obtained. Note that FIG. 4(a) shows the relationship between variables whose causal coefficient is not 0 (there is a causal relationship).

[0019] When expressing the causal exploration result in FIG. 4(a) as a causal graph, it becomes as shown in FIG. 4(b). In FIG. 4(b), the arrow indicates the presence of a causal relationship and the direction of causality, and w indicates the causal coefficient. Note that in FIG. 4(b), variables that are not recognized to have a causal relationship with other variables are not displayed. The causal exploration unit 14 passes the causal graph in FIG. 4(b) to the integration processing unit 18. On the other hand, the causal exploration unit 14 passes the causal exploration result in FIG. 4(a) to the effect manifestation delay estimation unit 16 as it is. Note that the causal exploration result in FIG. 4(a) can also be expressed as causal relationship data as shown in FIG. 5. Note that the causal relationship data in FIG. 5 is obtained by decomposing the causal graph in FIG. 4(b) (such that no arrow is connected from a plurality of variables to one variable).

[0020] Returning to FIG. 2, for a pair of two variables determined to have a causal relationship, the effect manifestation delay estimation unit 16 performs a cross-correlation analysis while shifting the time, and sets the timing at which the correlation becomes large as the effect manifestation delay value. Note that this effect manifestation delay value is referred to as the "delay time" hereinafter. In this cross-correlation analysis, for example, the technique disclosed in Japanese Patent Application Laid-Open No. 2021-9696 can be used. Specifically, as shown in FIG. 6(a), when there is data of the explanatory variable (data indicating that capital was invested in the 0th year) and data of the target variable (data indicating the improvement effect of the corporate value), the correlation between the target variable and the explanatory variable is calculated while shifting the year. Then, the effect manifestation delay estimation unit 16 sets the year in which the correlation value becomes maximum (year 6 in FIG. 6(a)) as the delay time.

[0021] The effect manifestation delay estimation unit 16 estimates the delay time for pairs of two variables (all pairs connected by arrows) for which it is determined that there is a causal relationship as shown in FIG. 5. That is, it estimates the delay time from the timing when there is an intervention (such as capital investment) in one of the variables in the pair until the effect appears in the other variable. Then, as shown in FIG. 6(b), the effect manifestation delay estimation unit 16 creates a causal search result (including delay information). In the causal search result (including delay information) shown in FIG. 6(b), the delay time is associated with pairs of causally related variables. Note that when reflecting the delay time in FIG. 6(b) in the causal relationship data in FIG. 5 (expressing the delay time), the causal relationship data (including delay information) shown in FIG. 7 is obtained. In FIG. 7, the arrow indicates that there is a causal relationship and the direction of causation, and the length of the arrow represents the delay time. Also, in FIG. 7, t, t - 1,... represent time. The effect manifestation delay estimation unit 16 outputs the causal relationship data (including delay information) in FIG. 7 to the integration processing unit 18.

[0022] As shown in FIG. 8, the integration processing unit 18 integrates the output result of the causal search unit 14 (FIG. 4(b)) and the output result of the effect manifestation delay estimation unit 16 (FIG. 7), and reconstructs a causal graph (causal tree) having the delay time between causally related variables as an attribute value. Specifically, the integration processing unit 18 reflects the delay time of the causal relationship data (including delay information) (the upper right figure in FIG. 8) in the tree structure of the causal graph (the upper left figure in FIG. 8), and creates a causal graph (after reconstruction) as shown in the lower figure in FIG. 8.

[0023] The output unit 20 displays the causal graph (after reconstruction) shown in FIG. 8 on the display unit 193.

[0024] (Flowchart) FIG. 9 is a flowchart showing the processing of the analysis device 10.

[0025] When the process of FIG. 9 starts, first, in step S10, the input reception unit 12 waits until there is an input of input information. When an analyst inputs input information (FIG. 3) via the input unit 195, the input reception unit 12 shifts to step S12, receives the input information, and passes it to the causal relationship exploration unit 14.

[0026] Next, in step S14, the causal relationship exploration unit 14 executes a causal relationship exploration process for calculating the presence or absence and direction of the causal relationship between variables. As a result, a causal relationship exploration result as shown in FIG. 4(a) is obtained. The causal relationship exploration unit 14 passes the causal relationship exploration result of FIG. 4(a) (the causal relationship data of FIG. 5) to the effect manifestation delay estimation unit 16, and at the same time, passes the causal graph (FIG. 4(b)) obtained from the causal relationship exploration result of FIG. 4(a) to the integration processing unit 18.

[0027] Next, in step S16, the effect manifestation delay estimation unit 16 executes an effect manifestation delay estimation process. That is, for each pair of variables that are considered to have a causal relationship in FIG. 5, cross-correlation analysis is performed while shifting the time, and the timing at which the correlation increases is set as the effect manifestation delay value for each pair. As a result, a causal relationship exploration result (including delay information) as shown in FIG. 6(b) and causal relationship data (including delay information) as shown in FIG. 7 are obtained. Therefore, the effect manifestation delay estimation unit 16 passes these data to the integration processing unit 18.

[0028] Next, in step S18, the integration processing unit 18 executes reconstruction of the causal graph. Specifically, as shown in FIG. 8, the causal graph received from the causal relationship exploration unit 14 and the causal relationship data (including delay information) received from the effect manifestation delay estimation unit 16 are integrated to create a causal graph (after reconstruction).

[0029] Next, in step S20, the output unit 20 displays (outputs) the reconstructed causal graph (Fig. 8) on the display unit 193. By referring to the causal graph (after reconstruction) shown in the lower diagram of Fig. 8, the analyst can confirm not only the causal relationship between the target variable and the first explanatory variable, but also the causal relationships among all variables including the second explanatory variable and the delay in the manifestation of the causal relationship. That is, from Fig. 8, it is possible to confirm the existence of the second explanatory variable (mediating variable) existing between the first explanatory variable and the target variable, and also to confirm the timing at which the effect spreads to the second explanatory variable. As a result, the analyst can explain the validity and reasons for the causal relationship from the explanatory variable to the target variable. In addition, it becomes easier to measure the effect from the capital investment in the explanatory variable until the effect spreads to the target variable, and it becomes possible to take additional measures at an early stage. Also, in Fig. 8, the connection of the causal relationship connecting from one first explanatory variable to the target variable is called a path, and the product of the causal coefficients included in this path (called the causal path coefficient) indicates the magnitude of the causal effect. For example, in the case of y (target variable) ← v1 ← x2 (first explanatory variable), the causal path coefficient = w y,v1 × w v1,x2 becomes. By comparing these causal path coefficients, the analyst can identify which first explanatory variable has a greater effect on the target variable.

[0030] (Regarding the application example) Next, the application example of the analysis apparatus 10 of the present embodiment will be described in detail.

[0031] Fig. 10 is a diagram showing an example of input information in this application example. In this application example, the target variable is the "sales revenue" of the company. Also, the first explanatory variables known to be related to the target variable are the "defect rate", "CO 2 emission amount", "work-life balance", etc. Also, the second explanatory variables whose relationships with the target variable and the first explanatory variables are unknown are the "warming index", "employee engagement", "human resource development", etc.

[0032] FIG. 11 shows an example of a causal exploration result obtained by performing causal exploration processing (S14) using the input information of FIG. 10. FIG. 12 shows a causal graph based on the causal exploration result of FIG. 11.

[0033] Furthermore, FIG. 13 shows an example of a causal exploration result (including delay information) obtained as a result of performing effect manifestation delay estimation processing (S16) based on the causal exploration result of FIG. 11. As a result of integrating the causal exploration result (including delay information) of FIG. 13 and the causal graph of FIG. 12, as shown in FIG. 14(a), a causal graph (after reconstruction) is obtained. In the case of this application example, the causal graph (after reconstruction) shown in FIG. 14 is output on the display unit 193.

[0034] According to this application example, it is possible to output a causal graph that can explain how the effects of capital investment in basic items (first explanatory variables) such as finance and non-finance in corporate management spread to the target variables (materiality, ESG indicators, corporate value, etc.). This causal graph can be used for management decision-making and can also be used as evidence in the financial statements for various stakeholders and analysts. FIG. 14(b) shows a comparative example. The comparative example in FIG. 14(b) is an example when the second explanatory variable is not input. According to this comparative example, since there is no second explanatory variable (mediating variable) between the target variable and the first explanatory variable, it is not possible to confirm in detail how the situation is until the effect appears in the target variable (sales revenue). On the other hand, by inputting the second explanatory variable (mediating variable) as in this application example, as shown in FIG. 14(a), it can be seen that the effect spreads to the second explanatory variable due to the intervention (capital investment) in the first explanatory variable. Therefore, by confirming the spread to the second explanatory variable before the effect of the capital investment in the first explanatory variable spreads to the target variable, it is possible to confirm whether the effect of the capital investment in the first explanatory variable has occurred.

[0035] Note that the application example is merely an example. Therefore, it is possible to apply this embodiment in fields other than corporate management.

[0036] As described in detail above, according to this embodiment, in the causal relationship exploration process (S14), the causal relationship exploration unit 14 identifies the causal relationship between one or more second explanatory variables different from the target variable and the first explanatory variable, and the first explanatory variable. Further, the effect manifestation delay estimation unit 16 estimates the delay time (delay in the occurrence timing of causality) in each causal relationship identified by the causal relationship exploration unit 14 in the effect manifestation delay estimation process (S16). Then, the integration processing unit 18 integrates the causal graph (FIG. 4(b)), which is the processing result of the causal relationship exploration unit 14, and the causal relationship data (including delay information) (FIG. 7), which is the processing result of the effect manifestation delay estimation unit 16, to reconstruct the causal graph (FIG. 8). Further, the output unit 20 outputs the causal graph (after reconstruction) onto the display unit 193. Thereby, in this embodiment, it is possible to provide the analyst with a causal graph that makes it easier to explain the relationship between the target variable and the first explanatory variable.

[0037] Here, in this embodiment, the analyst may input, as the second explanatory variable, an explanatory variable whose effect on the target variable is unknown. For this reason, the analyst does not have to understand and prepare all the explanatory variables, so the burden on the analyst can be reduced.

[0038] Further, according to this embodiment, the causal coefficient is displayed in the causal graph (after reconstruction). Thereby, the analyst can easily confirm which variables have a strong or weak causal relationship with which other variables.

[0039] (Modification Example 1) Hereinafter, Modification Example 1 will be described. In Modification Example 1, the integration processing unit 18 calculates the causal path coefficient of each path included in the causal graph (after reconstruction) in the lower diagram of FIG. 8, and changes the display order of each path based on the magnitude of the causal path coefficient. As described above, the causal path coefficient means the product of the causal coefficients included in one path. For example, in the causal graph (after reconstruction) of FIG. 8, the causal path coefficient w y,x1 of the path connecting x1 and y is the largest, and the causal path coefficient w y,x3is the second largest, and the causal path coefficient w of the path connecting x4 and y y,v2 × w v2,x4 is the third largest…, and so on. In this case, the integration processing unit 18 changes the display order of each path in the causal graph (after reconstruction) as shown in FIG. 15. By doing so, it becomes easier to recognize which first explanatory variable has a high or low effect on the target variable.

[0040] Note that the integration processing unit 18 may change the display mode such as the thickness and color of the arrows in FIG. 8 based on the causal coefficient and the causal path coefficient.

[0041] (Modification Example 2) Hereinafter, Modification Example 2 will be described. In Modification Example 2, the integration processing unit 18 changes the display order of each path in the causal graph (after reconstruction) in the lower diagram of FIG. 8 based on the delay time (the time until the intervention in the first explanatory variable affects the target variable). For example, in the case of the causal graph (after reconstruction) in FIG. 8, the delay time of the path connecting x1 and y is the shortest, the delay time of the path connecting x2 and y is the second shortest, and the delay time of the path connecting x4, x8 and y is the next shortest…, and so on. In this case, the integration processing unit 18 changes the display order of each path in the causal graph (after reconstruction) as shown in FIG. 16. By doing so, it becomes easier to recognize from which first explanatory variable the effect on the target variable spreads quickly.

[0042] (Modification Example 3) Hereinafter, Modification Example 3 will be described. In Modification Example 3, the integration processing unit 18 calculates the causal path coefficient of each path in the causal graph (after reconstruction) in the lower diagram of FIG. 8, calculates the sum of the causal path coefficients of all paths, and uses this as the causal coefficient of the target variable y. Then, the integration processing unit 18 displays the calculated causal coefficient of the target variable y together with the causal graph (after reconstruction).

[0043] For example, in the case of the causal graph (after reconstruction) in the lower diagram of FIG. 8, the causal coefficient of y is Causal coefficient of y = w y,x1 + w y,v1 × w v1,x2 + wy,x3 +… It can be calculated from an expression such as this. The causal coefficient of this y means that the greater the value, the greater the effect of the intervention on the first explanatory variable. Therefore, the analyst can estimate the total effect on the target variable y by referring to the causal coefficient of the target variable y.

[0044] Note that Modification 3 can be applied not only to the above-described embodiment but also to Modifications 1 and 2.

[0045] In the above-described embodiment and modifications, the case where the first explanatory variable and the second explanatory variable are separately input has been described (see FIG. 3), but it is not limited thereto. For example, as shown in FIG. 17, it may be input without distinguishing the explanatory variables. Also in this case, since the causal search unit 14 performs causal search using the target variable and the explanatory variables, it can identify the causal relationships between the variables as shown in FIG. 4(a). Therefore, finally, a causal graph (after reconstruction) as shown in FIG. 8 can be obtained. In this way, even if the explanatory variables are not distinguished into the first explanatory variable and the second explanatory variable, a causal graph that makes it easier to explain the relationship between the target variable and the explanatory variables can be provided to the analyst. That is, since the analyst does not have to understand all the explanatory variables, the burden on the analyst can be reduced.

[0046] Note that the above processing functions can be realized by a computer. In that case, a program that describes the processing contents of the functions that the processing device should have is provided. By executing that program on a computer, the above processing functions are realized on the computer. The program that describes the processing contents can be recorded on a computer-readable storage medium (excluding carrier waves).

[0047] When distributing a program, for example, it is sold in the form of a portable storage medium such as a DVD (Digital Versatile Disc) or a CD-ROM (Compact Disc Read Only Memory) on which the program is recorded. Also, the program can be stored in the storage device of a server computer and transferred from the server computer to other computers via a network.

[0048] A computer that executes a program stores, for example, a program recorded on a portable storage medium or a program transferred from a server computer in its own storage device. Then, the computer reads the program from its own storage device and executes processing according to the program. Note that the computer can also directly read the program from the portable storage medium and execute processing according to the program. Also, each time a program is transferred from a server computer, the computer can sequentially execute processing according to the received program.

[0049] The above-described embodiments are preferred examples of the present invention. However, the present invention is not limited thereto, and various modifications can be made without departing from the gist of the present invention.

[0050] In addition, the following supplementary notes are disclosed regarding the description of the above embodiments and modification examples. (Supplementary Note 1) Receiving an input of an objective variable and an explanatory variable, executing causal exploration using the objective variable and the explanatory variable to identify the causal relationship between each variable, estimating the delay in the occurrence timing of causality in each of the identified causal relationships, creating and outputting a causal graph representing the identified causal relationship and the estimated delay. A causal relationship analysis method characterized in that a computer executes the processing. (Appendix 2) The explanatory variable includes a first explanatory variable known to have a causal relationship with the target variable, and one or more second explanatory variables different from the first explanatory variable. The causal relationship analysis method according to Appendix 1 is characterized by this. (Appendix 3) In the specifying process, an index representing the strength of the specified causal relationship is calculated, The calculated index is displayed on the causal graph. The causal relationship analysis method according to Appendix 1 or 2 is characterized by this. (Appendix 4) In the specifying process, an index representing the strength of the specified causal relationship is calculated, In the process of creating and outputting the causal graph, the display order of the specified causal relationship is determined based on the magnitude of the calculated index. The causal relationship analysis method according to Appendix 1 or 2 is characterized by this. (Appendix 5) In the specifying process, an index representing the strength of the specified causal relationship is calculated, When there are a plurality of the first explanatory variables having a causal relationship with the target variable, the product of the indices indicating the strength of the causal relationship between variables existing between the target variable and each of the first explanatory variables is used as a value indicating the effect of each of the first explanatory variables on the target variable. Based on the value indicating the effect, the display order of the specified causal relationship is determined. The causal relationship analysis method according to Appendix 2 is characterized by this. (Appendix 6) In the process of creating and outputting the causal graph, the display order of the specified causal relationship is determined based on the magnitude of the delay estimated in the estimating process. The causal relationship analysis method according to Appendix 1 or 2 is characterized by this. (Appendix 7) In the specifying process, an index representing the strength of the specified causal relationship is calculated, When there are a plurality of the first explanatory variables having a causal relationship with the target variable, the product of the indices indicating the strength of the causal relationship between variables existing between the target variable and each of the plurality of first explanatory variables is calculated, and the sum of the calculated products is displayed as a value indicating the effect of the plurality of first explanatory variables on the target variable. The causal relationship analysis method according to any one of Appendices 1 to 5 is characterized by this. (Appendix 8) Accept the input of the target variable and the explanatory variable, Execute causal exploration using the target variable and the explanatory variables to identify the causal relationships between the variables, Estimate the delay in the timing of causation for each of the identified causal relationships, Create and output a causal graph representing the identified causal relationships and the estimated delays, A causal relationship analysis program, characterized by causing a computer to execute the processing. (Appendix 9) The explanatory variables include a first explanatory variable known to have a causal relationship with the target variable, and one or more second explanatory variables different from the first explanatory variable. The causal relationship analysis program according to Appendix 8, characterized by this.

Explanation of Signs

[0051] 10 Analysis device 12 Input reception unit 14 Causal exploration unit 16 Effect manifestation delay estimation unit 18 Integration processing unit 20 Output unit

Claims

1. accept the input of the target variable and the explanatory variable, perform causal exploration using the target variable and the explanatory variable to identify the causal relationship between each variable, estimate the delay in the timing of the occurrence of causality in each of the identified causal relationships, create and output a causal graph expressing the identified causal relationship and the estimated delay, A causal relationship analysis method characterized in that a computer executes the process.

2. The explanatory variable includes a first explanatory variable known to have a causal relationship with the target variable, and one or more second explanatory variables different from the first explanatory variable. The causal relationship analysis method according to Claim 1.

3. In the process of identifying, calculate an index representing the strength of the identified causal relationship, Display the calculated index on the causal graph. The causal relationship analysis method according to Claim 1 or 2, characterized in that.

4. In the process of identifying, calculate an index representing the strength of the identified causal relationship, In the process of creating and outputting the causal graph, determine the display order of the identified causal relationships based on the magnitude of the calculated index. The causal relationship analysis method according to Claim 1 or 2, characterized in that.

5. In the process of identifying, calculate an index representing the strength of the identified causal relationship, When there are a plurality of the first explanatory variables that have a causal relationship with the target variable, the product of the indexes indicating the strength of the causal relationship between the variables existing between the target variable and each of the first explanatory variables is used as a value indicating the effect of each of the first explanatory variables on the target variable, and based on the value indicating the effect, determine the display order of the identified causal relationships. The causal relationship analysis method according to Claim 2, characterized in that.

6. In the process of creating and outputting the causal graph, determine the display order of the identified causal relationships based on the magnitude of the delay estimated in the process of estimating. The causal relationship analysis method according to Claim 1 or 2, characterized in that.

7. In the process of identifying, calculate an index representing the strength of the identified causal relationship, When there are a plurality of the first explanatory variables that have a causal relationship with the target variable, calculate the product of the indices indicating the strength of the causal relationship between the target variable and each of the plurality of the first explanatory variables, and display the sum of the calculated products as a value indicating the effect of the plurality of the first explanatory variables on the target variable. The causal relationship analysis method according to claim 2, characterized in that.

8. Accept the input of the target variable and the explanatory variable, Execute causal exploration using the target variable and the explanatory variable to identify the causal relationship between each variable, Estimate the delay in the timing of the occurrence of causality in each of the identified causal relationships, Create and output a causal graph representing the identified causal relationship and the estimated delay. A causal relationship analysis program characterized by causing a computer to execute the process.

9. The explanatory variable includes a first explanatory variable known to have a causal relationship with the target variable and one or more second explanatory variables different from the first explanatory variable. The causal relationship analysis program according to claim 8, characterized in that.

Images