Analytical device and control method

Abstract

The analysis device (2000) obtains an object data set (10) and division number information (30), the object data set (10) being a set of object data (20) representing a pair of factor data (22) and result data (24). The division number information (30) represents the division number, which is the number of divisions used to divide the numerical range of the factor data (22). The analysis device (2000) calculates a plurality of boundary values, the plurality of boundary values ​​being used to divide the numerical range of the factor data (22) into the same number of ranges as the division number. In addition, the analysis device (2000) calculates a test statistic for each boundary value, the test statistic representing the difference associated with the result data (24) of two samples to be obtained by dividing the object data set (10) into two parts at the boundary value. Then, the analysis device (2000) generates test statistic information (40), the test statistic information representing a plurality of test statistics calculated for each boundary value.

Description

Technical Field

[0001] The present invention relates to data analysis. Background Art

[0002] A system for analyzing the content of a company's business has been developed. For example, Patent Document 1 discloses a device for determining whether there is an abnormality in a product by testing. More specifically, the device of Patent Document 1 determines whether there is an abnormality by performing a comparison test between a target value and an actual value for a variable representing a final quality characteristic of the product. Then, in the case where the final quality characteristic includes an abnormality, the device of Patent Document 1 also determines whether there is an abnormality by performing a comparison test between a target value and an actual value for other variables associated with the final quality characteristic.

[0003] Prior art literature

[0004] Patent Literature

[0005] [PTL 1] Japanese Patent Application Publication No. 2019-36061 Summary of the invention

[0006] Technical issues

[0007] The premise of Patent Document 1 is that a target value indicating the value of an appropriate variable is known in advance. Therefore, the device in PTL 1 is not necessarily used to determine the target value of the variable.

[0008] The present invention has been conceived in view of the above-mentioned problems, and one of its objects is to provide a technique that facilitates identification of appropriate values ​​of factors that may affect results.

[0009] Solution to the problem

[0010] The analysis device according to the present invention includes: (1) an acquisition unit, which acquires a set of object data and division quantity information, each object data indicates a pair of factor data and result data, the factor data is a value related to the factor, the result data is a value related to the result, and the division quantity information indicates the number of numerical ranges used to divide the factor data; (2) a calculation unit, which calculates a test statistic for each boundary value of a plurality of boundary values, the plurality of boundary values ​​are used to divide the numerical range into quantities, the test statistic represents a difference associated with result data of two samples to be obtained by dividing the object data included in the set of object data into two parts at the boundary value; and (3) a generation unit, which generates test statistic information, the test statistic information indicating a plurality of test statistics calculated for each boundary value.

[0011] The control method according to the present invention is executed by a computer. The control method includes: (1) an acquisition step of acquiring a set of object data and division quantity information, each object data indicating a pair of factor data and result data, the factor data being a value associated with the factor, the result data being a value associated with the result, and the division quantity information indicating the number of numerical ranges for dividing the factor data; (2) a calculation step of calculating a test statistic for each boundary value of a plurality of boundary values, the plurality of boundary values ​​being used to divide the numerical range into quantities, the test statistic representing a difference associated with result data of two samples to be obtained by dividing the object data included in the set of object data into two parts at the boundary value; and (3) a generation step of generating test statistic information, the test statistic information indicating a plurality of test statistics calculated for each boundary value.

[0012] The program according to the present invention causes a computer to execute the control method according to the present invention.

[0013] Advantageous Effects of the Invention

[0014] Techniques are provided that facilitate identifying appropriate values ​​for factors that may affect the results. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a schematic diagram showing an overview of an analysis device of an exemplary embodiment of the present invention.

[0016] Figure 2 is a diagram conceptually illustrating the process of calculating the test statistic for each sample pair.

[0017] Figure 3 is a schematic diagram showing a functional configuration of an analysis device according to Exemplary Embodiment 1.

[0018] Figure 4 is a schematic diagram showing a computer for realizing the analysis device.

[0019] Figure 5 is a schematic diagram showing the usage environment of the analysis device.

[0020] Figure 6 is a flowchart showing the flow of processing to be performed by the analysis device according to Exemplary Embodiment 1.

[0021] Figure 7 is a schematic diagram showing test statistic information to be implemented as a table.

[0022] Figure 8 is a schematic diagram showing test statistic information to be implemented as a test statistic graph.

[0023] Fig. 9 is a schematic diagram showing a screen to be provided to the user.

[0024] Fig.10 is a diagram showing a screen including a histogram and test statistic information.

[0025] Fig.11 is a schematic diagram showing a screen including a time series graph, a histogram, and a test statistic graph.

[0026] Fig.12 is a schematic diagram illustrating a first sample and a second sample in an analysis device 2000 according to a modification.

[0027] Fig.13 is a schematic diagram showing another method of dividing the object data set 10 into two samples using a pair of two boundary values.

[0028] Fig.14 is a schematic diagram showing information to be output by an analysis device 2000 according to a modification. DETAILED DESCRIPTION

[0029] The following uses the accompanying drawings to describe exemplary embodiments of the present invention. Note that in all the accompanying drawings, similar reference numerals are used to indicate similar constituent elements, and their descriptions are omitted as appropriate. In addition, in each block diagram, each block represents a configuration based on function rather than a configuration based on hardware, unless otherwise described.

[0030] [Exemplary Embodiment 1]

[0031] <Overview>

[0032] Figure 1 is a schematic diagram showing an overview of the analysis device 2000 of the present exemplary embodiment. Note that Figure 1 This is a diagram for the purpose of explaining the analysis device 2000 to facilitate understanding. The functions of the analysis device 2000 are not limited to Figure 1 Functions shown.

[0033] The analysis device 2000 acquires an object data set 10 which is a set of a plurality of object data 20. The object data 20 is data in which a factor value (factor data 22) and a result value (result data 24) are associated with each other. For example, when manufacturing a product, the content of a specific material affects the quality of the product. In view of this, it is conceived to process a flag indicating the content of the specific material (as the factor data 22) and indicating whether the product is a defective product (as the result data 24).

[0034] The analysis device 2000 divides each object data 20 into two parts by a plurality of different patterns, and calculates a test statistic for each pattern, the test statistic representing the difference between two sets (samples) to be obtained by dividing into two parts. To this end, the analysis device 2000 further acquires the division number information 30. The division number information 30 indicates the number of divisions into which the numerical range of the factor data 22 is divided (hereinafter referred to as the division number).

[0035] The analysis device 2000 calculates each boundary value for dividing the numerical range of the factor data 22 into the number of divisions indicated by the division number information 30. For example, it is assumed that the numerical range of the factor data 22 is from 0 to 100, and the number of divisions is 4. The boundary values ​​for equally dividing the numerical range into four parts are 25, 50, and 75. Therefore, these three boundary values ​​are determined by the analysis device 2000.

[0036] The analysis device 2000 performs the following processing on each of the determined multiple boundary values. First, the analysis device 2000 divides the object data set 10 into two sets (hereinafter referred to as the first sample and the second sample) using the boundary value as the boundary. Therefore, pairs of the first sample and the second sample (hereinafter referred to as sample pairs) are obtained, and the number of pairs is equal to the number of divisions.

[0037] In addition, for each boundary value, the analysis device 2000 calculates a test statistic representing a difference related to the result data 24 for the sample pairs obtained by the division based on the boundary value. For example, it is assumed that the result data 24 is a defect flag. In this case, for example, the analysis device 2000 calculates a test statistic representing the difference between the defect rate (the ratio of defective products to the total number of products) of the products in the first sample and the defect rate of the products in the second sample.

[0038] Figure 2 is a diagram conceptually illustrating the process of calculating the test statistic for each sample pair. Figure 2 The graph in indicates the value of the factor data 22 on the horizontal axis and the value of the result data 24 on the vertical axis. The result data 24 indicates the defect rate. In this example, the number of divisions is 4. Therefore, three boundary values ​​b1, b2, and b3 are calculated. Note that min and max on the horizontal axis are the upper limit and lower limit of the numerical range of the factor data 22, respectively.

[0039] The analysis device 2000 divides the object data set 10 into a first sample and a second sample by using a boundary value b1, the first sample including object data 20 in which one factor data 22 is equal to or less than bl, and the second sample including object data 20 in which one factor data 22 is greater than bl. Then, the analysis device 2000 calculates a test statistic representing the difference associated with the result data 24 of the first sample and the second sample as a test statistic associated with the boundary value b1. For example, in the case where the result data 24 indicates whether the product is a defective product, it is conceived to calculate the defect rate of each sample of the first sample and the second sample, and calculate a test statistic representing the difference between the calculated defect rates. Similarly, the analysis device 2000 also calculates a test statistic representing the difference associated with the result data 24 of the first sample and the second sample to be obtained by using the boundary value b2, and a test statistic representing the difference associated with the result data 24 of the first sample and the second sample to be obtained by using the boundary value b3.

[0040] The analysis device 2000 generates test statistic information indicating a plurality of calculated test statistics. For example, the test statistic information is a graph indicating the correlation between the boundary value and the sample pairs obtained by dividing with reference to the boundary value. Figure 1 In the example, a line graph is generated as the test statistic information 40, in which the boundary value is indicated on the horizontal axis and the test statistic calculated at the boundary value is indicated on the vertical axis.

[0041] <An example of beneficial effect>

[0042] In the analysis device 2000 according to the present exemplary embodiment, the process of "dividing the object data set 10 into two samples based on the value of one factor data 22, and calculating the test statistic representing the difference related to the result data 24 of the two samples" is performed through multiple modes, each mode using a different boundary value. Then, the test statistic information 40 representing the test statistic obtained for each mode of the multiple modes is output. A user who views the test statistic information 40 as described above can easily identify "what value of one factor data 22 is suitable for obtaining a good result". Specifically, the analysis device 2000 can accurately and easily identify what value of one factor data 22 is appropriate in terms of the relationship with the result data 24.

[0043] Here, advanced statistical knowledge is required to appropriately perform data analysis. Therefore, when a person other than an expert with such knowledge attempts to perform data analysis, a large amount of time may be required, or errors may occur in the analysis. In this regard, the analysis device 2000 automatically performs the process of dividing the data by a plurality of patterns, and the process of calculating the test statistic for each pattern, so that even a person who does not have advanced statistical knowledge (e.g., a business person in a company, etc.) can accurately and easily identify the appropriate value of the factor data 22 that may affect the result data 24.

[0044] This exemplary embodiment is described in further detail below.

[0045] <Function configuration example>

[0046] Figure 3 2 is a schematic diagram showing a functional configuration of an analysis device 2000 according to exemplary embodiment 1. The analysis device 2000 includes an acquisition unit 2020, a calculation unit 2040, and a generation unit 2060. The acquisition unit 2020 acquires the object data set 10 and the division number information 30. The calculation unit 2040 calculates a plurality of boundary values ​​for dividing the numerical range of the factor data 22 into the number of divisions indicated by the division number information 30. In addition, for each of the plurality of boundary values, the calculation unit 2040 calculates a test statistic representing a difference associated with result data 24 of a first sample and a second sample to be obtained by dividing the object data set 10 into two parts at the boundary value. The generation unit 2060 generates test statistic information indicating a plurality of calculated test statistics.

[0047] <Hardware Configuration Example of Analysis Device 2000>

[0048] Each functional configuration unit of the analysis device 2000 may be implemented by hardware for implementing each functional configuration unit (for example, a hard-wired electronic circuit, etc.), or may be implemented by a combination of hardware and software (for example, a combination of an electronic circuit and a program for controlling the electronic circuit, etc.). The following further describes the case where each functional configuration unit of the analysis device 2000 is implemented by a combination of hardware and software.

[0049] Figure 4 is a schematic diagram showing a computer 1000 for realizing the analysis device 2000. The computer 1000 is any computer. For example, the computer 1000 is a fixed computer such as a personal computer (PC) and a server machine. In addition, for example, the computer 1000 is a portable computer such as a smartphone or a tablet terminal.

[0050] The computer 1000 may be a dedicated computer designed to realize the analyzing apparatus 2000, or may be a general-purpose computer. In the latter case, for example, the computer 1000 realizes each function of the analyzing apparatus 2000 by installing a predetermined application in the computer 1000. The above-mentioned application is configured as a program that realizes the functional configuration unit of the analyzing apparatus 2000.

[0051] The computer 1000 includes a bus 1020, a processor 1040, a memory 1060, a storage device 1080, an input / output interface 1100, and a network interface 1120. The bus 1020 is a data transmission path through which the processor 1040, the memory 1060, the storage device 1080, the input / output interface 1100, and the network interface 1120 transmit and receive data to and from each other. However, the method of connecting the processor 1040 and the like to each other is not limited to bus connection.

[0052] The processor 1040 is various processors such as a central processing unit (CPU), a graphics processing unit (GPU), and a field programmable gate array (FPGA). The memory 1060 is a main storage device implemented using a random access memory (RAM) or the like. The storage device 1080 is an auxiliary storage device implemented using a hard disk, a solid state drive (SSD), a memory card, a read-only memory (ROM), or the like.

[0053] The input / output interface 1100 is an interface that connects the computer 1000 to an input / output device. For example, the input / output interface 1100 is connected to an input device such as a keyboard and an output device such as a display device.

[0054] The network interface 1120 is an interface for connecting the computer 1000 to a communication network. The communication network is, for example, a local area network (LAN) or a wide area network (WAN).

[0055] The storage device 1080 stores a program (a program that realizes the above-mentioned application) for realizing each functional configuration unit of the analysis apparatus 2000. The processor 1040 realizes each functional configuration unit of the analysis apparatus 2000 by reading the program in the memory 1060 and executing the program.

[0056] <Example of Usage Environment of Analysis Device 2000>

[0057] In order to facilitate understanding of the analysis device 2000 , an example of a usage environment of the analysis device 2000 is described. Figure 5 2 is a schematic diagram showing a usage environment of the analysis device 2000 .

[0058] exist Figure 5In the example, the analysis device 2000 is connected to the user terminal 60 via a network. The user operates the user terminal 60 and transmits a request to the analysis device 2000 so that the test statistic information 40 about the object data set 10 can be provided. For example, the request includes information indicating a condition related to the object data set 10 and the partition number information 30. The analysis device 2000 acquires the object data set 10 associated with the condition indicated in the request, and generates the test statistic information 40 by using the object data set 10 and the partition number information 30. Then, the analysis device 2000 sends the generated test statistic information 40 to the user terminal 60.

[0059] For example, the analysis device 2000 provides the user terminal 60 with screen data (eg, a web page) including the test statistic information 40. In this case, the user terminal 60 displays the received web page through a browser. By doing so, the user can browse the test statistic information 40.

[0060] The use environment of the analysis device 2000 is not limited to Figure 5 For example, the analysis device 2000 may be directly operated by a user instead of being used via the user terminal 60 .

[0061] <Processing Flow>

[0062] Figure 6 2 is a flowchart showing a process flow to be performed by the analysis device 2000 according to the exemplary embodiment 1. The acquisition unit 2020 acquires the object data set 10 (S102). The acquisition unit 2020 acquires the division number information 30 (S104). The calculation unit 2040 calculates a plurality of boundary values ​​for dividing the numerical range of the factor data 22 into the number of divisions indicated by the division number information 30 (S106).

[0063] S108 to S114 are loop processing A to be performed for each boundary value. In S108, the calculation unit 2040 determines whether the loop processing A has been performed on all boundary values ​​as objects. In the case where the loop processing A has been performed on all boundary values ​​as objects, Figure 6 The processing in proceeds to S116. On the other hand, in the case where there are boundary values ​​that have not yet been subjected to loop processing A as an object, the calculation unit 2040 selects one of the boundary values. Then, Figure 6 The processing in proceeds to S110. Note that the boundary value selected here is referred to as boundary value i.

[0064] The calculation unit 2040 generates a first sample and a second sample by dividing the object data set 10 into two parts at the boundary value i (S110). The calculation unit 2040 calculates the test statistic related to the result data 24 of the first sample and the second sample (S112). Since S114 is the termination of the loop process A, Figure 6 The processing proceeds to S108.

[0065] After the loop process A, the generation unit 2060 generates the test statistic information 40 for the plurality of calculated test statistics (S116). The generation unit 2060 outputs the test statistic information 40 (S118).

[0066] <About object data set 10>

[0067] The object data set 10 is a set including a plurality of object data 20, each of which is a pair of factor data 22 and result data 24. Each piece of object data 20 indicates data, for example, for each case where the same process or the same operation is performed at different times, respectively. For example, as described above, it is assumed that the object data 20 indicates the content of a specific material and whether a product manufactured by using the material containing the content is a defective product for each of the factor data 22 and the result data 24. In this case, for example, for products manufactured at different times, respectively, each piece of object data 20 indicates the content of the material having the above-mentioned characteristics and used in the product manufactured at this time and whether the product is a defective product.

[0068] Here, learning data for learning an estimation model for estimating the value of a target variable from the values ​​of a plurality of explanatory variables may be used as the object data 20. For example, it is assumed that an estimation model is generated that estimates an index related to a product to be produced in a manufacturing condition based on each index one level higher than an index representing a product manufacturing condition. In this case, each index representing a manufacturing condition represents an explanatory index, and an index related to a manufactured product represents a target variable. The index representing the manufacturing condition is, for example, an environmental condition such as temperature and humidity, the content of each material, and the like. The index related to a product is, for example, a flag indicating whether a product is a defective product, an index indicating a specific performance related to a product, and the like.

[0069] The values ​​of the explanatory variables and the values ​​of the objective variables described above can be respectively used as the factor data 22 and the result data 24. In view of this, for example, learning data to be used in the generation of such an estimation model can be used as the object data 20.

[0070] The acquisition unit 2020 acquires the object data set 10 (S102). There are various methods for acquiring the object data set 10 by the acquisition unit 2020. For example, the acquisition unit 2020 acquires the object data set 10 from a storage device storing the object data set 10. In addition, for example, the acquisition unit 2020 can acquire the object data set 10 by receiving the object data set 10 transmitted from another device.

[0071] Here, in the case where a plurality of object data sets 10 are stored in the storage device, the acquisition unit 2020 determines the object data set 10 from which the test statistic information 40 is to be generated, and acquires from the storage device the determined object data set 10. In this case, for example, the acquisition unit 2020 acquires a condition related to the object data set 10 to be acquired, and acquires the object data set 10 matching the condition.

[0072] The conditions related to the object data set 10 are, for example, identification information (e.g., name) of the object data set 10, a time period for acquiring the object data set 10, a location related to the object data set 10 (e.g., a manufacturing location of a related product, etc.), etc. Here, in the case where the object data set 10 is used to generate an estimation model, identification information of the estimation model generated using the object data set 10 may be used as identification information of the object data set 10. Information indicating these conditions related to the object data set 10 is included in, for example, a request to be sent from the above-mentioned user terminal 60.

[0073] Note that the object data 20 stored in the storage device is not necessarily divided in units of the object data set 10. In this case, for example, the acquisition unit 2020 acquires conditions related to the object data 20 included in the object data set 10, and acquires the object data 20 that meets the conditions from the storage device. Then, the acquisition unit 2020 processes the group of acquired object data 20 as the object data set 10. The conditions related to the object data 20 are, for example, a time period for acquiring the object data 20, a place related to the object data 20 (for example, a manufacturing place of a product), and the like.

[0074] In addition to this, for example, in the case where another process using the object data set 10 has been performed before generating the test statistic information 40, the analysis device 2000 can obtain the object data set 10 by taking over the object data set 10 from the other process as described above. For example, a process of generating an estimation model from the object data set 10, a process of visualizing the object data set 10 using a graph, etc., etc. are regarded as another process using the object data set 10.

[0075] For example, an interface (e.g., a button) for requesting the generation of test statistic information 40 related to the object data set 10 used in the process is provided on the screen showing the processing result. In response to the request using the interface, the analysis device 2000 generates and outputs the test statistic information 40. Therefore, for the object data set 10 used in generating the estimation model or the visualized object data set 10, the user can easily obtain the information of the so-called "test statistic information 40", and in other words, the information is obtained by analyzing the object data set 10.

[0076] Note that, when the device performing processing such as generating an estimation model is different from the analysis device 2000 , both the identification information of the object data set 10 and the object data set 10 itself are included in the request from the device.

[0077] Note that, in the case where the object data 20 includes data of each factor of a plurality of factors, it is necessary to determine which factor is to be taken as an object to divide the object data set 10 into samples (specifically, which data of the factor is to be processed as the factor data 22). The determination method is optional. For example, the acquisition unit 2020 receives an input from a user, which specifies which data of the factor is to be processed as the factor data 22. In addition to this, for example, in the case where the object data set 10 is taken over from another process as described above, the data of the factor taken as an object in the other process may be processed as the factor data 22. For example, in the case where the object data set 10 is taken over from a process of visualizing the data of a certain factor, the visualized data of the factor may be processed as the factor data 22.

[0078] <Acquisition of division quantity information 30: S104>

[0079] The acquisition unit 2020 acquires the division number information 30 (S104). There are various methods of acquiring the division number information 30 by the acquisition unit 2020. For example, the acquisition unit 2020 acquires the division number information 30 stored in the storage device. In addition to this, for example, the acquisition unit 2020 can acquire the division number information 30 by accepting an operation of inputting the division number by the user. In addition to this, for example, the acquisition unit 2020 can acquire it by receiving the division number information 30 transmitted from another device. In this case, for example, the division number information 30 is included in the request to be transmitted from the above-mentioned user terminal 60.

[0080] <Calculation boundary value: S106>

[0081] The calculation unit 2040 calculates a plurality of boundary values ​​(S106) for dividing the numerical range of the factor data 22 into the number of divisions indicated by the division number information 30. In view of this, the calculation unit 2040 determines the numerical range of the factor data 22. For example, the calculation unit 2040 processes, as the numerical range of the factor data 22, a numerical range in which the minimum value of the factor data 22 included in the object data set 10 is set as a lower limit and the maximum value of the factor data 22 included in the object data set 10 is set as an upper limit.

[0082] In addition to this, for example, the calculation unit 2040 may determine the numerical range of the factor data 22 after eliminating a part of the object data 20 included in the object data set 10. In this case, for example, the calculation unit 2040 processes the numerical range with the minimum value and the maximum value of the factor data 22 included in the object data set 10 after the elimination set as the lower limit and the upper limit, respectively, as the numerical range of the factor data 22. The object data 20 to be eliminated from the object data set 10 is, for example, data in which one of the factor data 22 indicates an abnormal value. Note that various existing methods can be used as a method of determining an abnormal value from a plurality of numerical values.

[0083] In addition to this, for example, the calculation unit 2040 may acquire information in which a numerical range of the factor data 22 is defined. In this case, a lower limit of the numerical range of the factor data 22 may be less than a minimum value of one factor data 22 included in the object data set 10, or an upper limit of the numerical range of the factor data 22 may be greater than a maximum value of one factor data 22 included in the object data set 10.

[0084] The calculation unit 2040 calculates each boundary value for dividing the numerical range of the factor data 22 into the number of divisions indicated by the division number information 30. For example, the boundary value can be calculated by the following equation (1).

[0085] b_i=(max-min) / N*i (1)

[0086] Here, bi represents the i-th boundary value counted from the minimum value. max represents the upper limit of the numerical range of the factor data 22, and min represents the lower limit of the numerical range of the factor data 22. N represents the number of divisions. Note that the number of boundary values ​​becomes N-1.

[0087] <Generate sample pair: S110>

[0088] By dividing the object data set 10 into two parts at the boundary value, the calculation unit 2040 generates a sample pair (a first sample and a second sample) for each boundary value (S110). For example, the calculation unit 2040 divides the object data set 10 into two parts, that is, a set of object data 20 indicating that the factor data 22 is equal to or less than the boundary value, and a set of object data 20 indicating that the factor data 22 is greater than the boundary value. The former is processed as the first sample, and the latter is processed as the second sample. However, the calculation unit 2040 can generate and set a set of object data 20 indicating that one factor data 22 is less than the boundary value as the first sample, and generate and set a set of object data 20 indicating that one factor data 22 is equal to or greater than the boundary value as the second sample.

[0089] <Calculate test statistic: S112>

[0090] The calculation unit 2040 calculates a test statistic for each boundary value, and the test statistic represents the difference related to the result data 24 of the first sample and the second sample generated using the boundary value (S112). Here, as a test for comparing the two samples as described above, a ratio difference test, a mean difference test, etc. can be used. The calculation unit 2040 calculates test statistics that can be used for these tests. As the category of the test statistic, various categories such as a two-sample t test can be used.

[0091] For example, it is assumed that a test on the difference in ratio is performed on the first sample and the second sample. In this case, the calculation unit 2040 calculates a test statistic representing the difference between the ratio of a predetermined category related to the result data 24 included in the first sample and the ratio of a predetermined category related to the result data 24 included in the second sample. For example, in the case where one result data 24 is a defect flag, as the ratio of the prescribed category related to the result data 24, a defect rate (where the factor data 22 indicates the ratio of the number of object data 20 of defective products relative to the total number of object data 20 included in the sample) can be used.

[0092] Note that the categories of tests and test statistics may be pre-defined or specified by user input or the like.

[0093] <Generate test statistic information: S116>

[0094] The generation unit 2060 generates test statistic information 40 indicating the test statistic calculated for each boundary value (S116). The test statistic information 40 indicates that the boundary value is associated with the test statistic calculated for the boundary value. As such information, various forms of information such as a graph can be adopted.

[0095] For example, the test statistic information 40 includes a table indicating the correlation between boundary values ​​and test statistics. Figure 7 is a schematic diagram showing test statistic information 40 to be implemented as a table. Figure 7 The table in is called table 50. Table 50 includes boundary values ​​52 and test statistics 54. Figure 7 In FIG. 5 , one record in table 50 indicates mutually associated boundary values ​​and a test statistic calculated for two samples obtained by partitioning the object data set 10 at the boundary values.

[0096] The test statistic information 40 may be a graph indicating the correlation between the boundary value and the test statistic (hereinafter referred to as a test statistic graph). Figure 8 is a schematic diagram showing the test statistic information 40 to be implemented as a test statistic graph 70. In the test statistic graph 70, the X-axis represents the boundary value, and the Y-axis represents the test statistic.

[0097] In the test statistic graph 70, each correlation between the boundary value and the test statistic is marked, and a line connecting these marks is displayed. Note that the type of graph is not limited to a line graph, and various graphs such as a bar graph may be employed.

[0098] The test statistic graph 70 includes a threshold indication 74 indicating a test statistic threshold. The threshold indication 74 indicates a threshold of the test statistic to be obtained based on the significance level. In the test statistic graph 70, the test statistic that is equal to or greater than the threshold indicated by the threshold indication 74 is a statistically valid value. Specifically, in the case where the test statistic exceeds the threshold, it can be said that there is a statistically significant difference between the two samples divided by the boundary value associated with the test statistic. In this way, by including the threshold indication 74 in the test statistic graph 70, it is possible to easily identify whether the test statistic associated with each boundary value is a statistically valid value.

[0099] For example, in Figure 8 In the example, it is assumed that a piece of result data 24 is a defect flag, and factor data 22 represents the content of component A. Furthermore, it is assumed that a test for the difference in defect rates between samples is performed. Furthermore, it is assumed that the test statistic has a positive value when the defect rate of the second sample is higher than that of the first sample.

[0100] Observing the test statistic graph 70, when the boundary value is equal to or less than b3, the test statistic becomes less than the threshold value. On the other hand, when the boundary value is equal to or greater than b4, the test statistic becomes greater than the threshold value. In view of this, when comparing products with a content of component A equal to or greater than b4 with products with a content of component A less than b4, there is a significant difference in the defect rate of the products. Therefore, when the user observes Figure 8When looking at the test statistic graph 70, it can be easily recognized that "in order to reduce the defect rate, it is appropriate to set the content of component A to less than b4".

[0101] Note that Figure 8 In the example, by giving a color (a pattern for convenience of explanation) to the area above the threshold indicator 74, it becomes easier to identify the area where the test statistic is equal to or greater than the threshold and the area other than the threshold. Coloring may also be performed in the area below the threshold indicator 74. In addition, coloring may not be performed in the area above the threshold indicator 74, or coloring may be performed only in the area below the threshold indicator 74.

[0102] In addition, the generation unit 2060 may display a mark indicating a test statistic equal to or greater than a threshold in an emphasized manner in the test statistic graph 70. As a method of emphasizing display, various methods such as increasing the size of the mark, changing the shape of the mark, and making the mark blink may be adopted.

[0103] By converting the significance level into the test statistic, the threshold value of the test statistic can be obtained. For example, assuming that the significance level is 5% and the category of the test statistic is the t value. In this case, by converting the p value = 0.05 representing the significance level of 5% into the t value, the threshold value of the t value can be obtained. Note that as a specific method of converting the value representing the significance level into the test statistic, an existing method according to the category of the test statistic can be used.

[0104] The significance level may be pre-defined or may be specified by the user. For example, the analysis device 2000 provides the user with a screen through which information such as the number of divisions, test categories, and significance level may be specified.

[0105] Fig. 9 is a schematic diagram showing a screen to be provided to the user. Fig. 9 The screen 80 in FIG. 8 includes input areas 82, 84, and 86. Each of these areas is an input area through which the category of the test, the significance level, and the number of partitions can be specified. Note that because the number of partitions can also be said to be the granularity of the test, Fig. 9 In the example, the number of divisions is indicated as the "test granularity". After entering information into these input areas, the user presses button 88 to perform the test. Figure 8 The test statistic graph 70 is shown. Note that the test statistic graph 70 can be displayed on the screen 80 or on a screen other than the screen 80. Note that the button 88 is not a necessary element, and the test statistic graph 70 reflecting the input content can be automatically generated and displayed in response to the input to each input area.

[0106] <Output test statistic information 40:118>

[0107] The generation unit 2060 outputs the generated test statistic information 40 (S118). The output destination of the test statistic information 40 is optional. For example, the generation unit 2060 causes the display device accessible from the analysis device 2000 to display the test statistic information 40. In addition, for example, the generation unit 2060 stores the test statistic information 40 in a storage device accessible from the analysis device 2000. In addition, for example, the generation unit 2060 transmits the test statistic information 40 to another device (e.g., user terminal 60) accessible from the analysis device 2000.

[0108] <Show Histogram>

[0109] The generation unit 2060 may generate a histogram related to the factor data 22 included in the object data set 10, and output the histogram together with the test statistic information 40. For example, the histogram and the test statistic information 40 may be output by including them in the same screen. Fig.10 is a schematic diagram showing a screen including a histogram 90 and test statistic information 40. Fig.10 , histogram 90 is included in the above-mentioned inspection statistic graph 70. Note that, as will be described below Fig.11 As shown, a test statistic graph 70 and a histogram 90 may also be displayed separately.

[0110] For example, the histogram 90 is generated as follows. First, the generating unit 2060 generates a local range by dividing the numerical range of the factor data 22 at each boundary value. The generating unit 2060 counts the object data 20 whose factor data 22 belongs to the local range for each local range. Then, the generating unit 2060 generates a histogram 90 representing the correlation between the local range and the number of object data 20.

[0111] In this way, by outputting the histogram 90 in addition to the test statistic information 40 , the user can accurately and easily identify what values ​​of the factor data 22 lead to better results, while understanding how the factor data 22 are distributed.

[0112] <More specific output example>

[0113] In addition to the above-mentioned information, the generation unit 2060 may also generate various information. Fig.11 is a schematic diagram showing a more specific example of information to be generated by the generating unit 2060. In this example, the screen 130 is generated by the generating unit 2060. The screen 130 includes a display area 131, a display area 132, a display area 133, a display area 134, and a display area 135.

[0114] A time series graph 100 showing the time change of the object data 20 is displayed in the display area 131. In the time series graph 100, a line 102 shows the time change of the factor data 22. In this case, for example, the object data set 10 is time series data in which each object data 20 is associated with time. However, by accumulating the data in a specific cycle, the data can be processed and used as time series data.

[0115] The histogram 90 is displayed in the display area 132. The input interface for specifying the test method and the test granularity (equivalent to the number of divisions) and the button for instructing to perform the test are displayed in the display area 133.

[0116] An input interface for specifying a significance level and a test statistic graph 70 are displayed in the display area 134 . Fig.11 The test statistic graph 70 in FIG. 7 includes a bar 72 and an explanation indicator 76. The bar 72 is an interface for specifying a desired boundary value. The user can change the desired boundary value by moving the bar 72 left and right. The explanation indicator 76 indicates the boundary value specified by the bar 72 and the test statistic of the boundary value.

[0117] The histogram 90 has different display schemes in the levels below the boundary value specified by the bar 72 and the levels above the boundary value. Specifically, different patterns are given to the data of each level whose value is less than the boundary value b3 specified by the bar 72 and the data of each level whose value is greater than the boundary value b3. Note that the display scheme to be distinguished is not limited to the pattern, and the category of color, the category of border line, etc. can be used.

[0118] Furthermore, a line 110 indicating the defect rate of each class is indicated on the histogram 90 in the display area 132. Furthermore, a confidence interval of the defect rate in each class is indicated by a dotted line.

[0119] The display area 135 indicates a simulation of the degree of improvement of the value related to the result data 24 based on the boundary value specified by the bar 72. For example, it is assumed that a test statistic is calculated for the defect rate. In addition, it is assumed that the defect rate is reduced by reducing the value of the factor data 22. In this case, the display area 135 includes a display area 136 and a display area 137, the display area 136 indicating the defect rate and the like in the entire object data 20, and the display area 137 indicating the defect rate and the like in the case where the factor data 22 is limited to a value equal to or less than the boundary value specified by the bar 72.

[0120] It can be said that the information indicated in the display area 137 indicates how much improvement can be expected by adjusting the factor (the content of component A in this example) so that the factor data 22 becomes equal to or less than the specified boundary value. Therefore, according to this display, the effect (improvement in the defect rate in this example) that can be expected by adjusting the factor can be easily recognized.

[0121] Here, as described above, the designated boundary value can be changed by allowing the user to move the bar 72 left and right. The generation unit 2060 changes the content of the display area 137 in response to the movement of the bar 72 (changing the designation of the boundary value). Therefore, the user can confirm the effect that can be expected by adjusting the factor while moving the bar 72.

[0122] Furthermore, the generation unit 2060 also changes the display scheme of the data in the histogram 90 in response to the movement of the bar 72. By doing so, the user can easily confirm the samples used in the simulation for improvement.

[0123] Note that Fig.11 In the example of , since the defect rate is reduced by reducing the factor data 22, the display area 137 displays "adjust to b3 or less". In this regard, in the case where the defect rate is reduced by increasing the factor data 22, the display changes to "adjust to b3 or more".

[0124] Information for determining whether to reduce or increase the factor data 22 for improvement is acquired in advance. For example, as described above, it is assumed that the learning data for learning the estimation model is used as the object data 20, and the estimation model estimates the value of the target variable based on the values ​​of a plurality of explanatory variables. In this case, it can be determined whether to reduce or increase the factor data 22 associated with a certain explanatory variable for improvement depending on what influence, positive influence or negative influence, the explanatory variable gives to the target variable in the estimation model.

[0125] [Modifications]

[0126] In the examples described so far, the object data 20 associates one piece of factor data 22 with one piece of result data 24. Alternatively, the object data 20 may associate two pieces of factor data 22 with one piece of result data 24. In this case, when the object data set 10 is divided into two samples, the division is performed by focusing on two factors. The analysis device 2000 that performs such processing is hereinafter referred to as the analysis device 2000 according to the modified example.

[0127] In the analysis device 2000 according to the variation, the acquisition unit 2020 acquires the object data set 10 including a plurality of object data 20 , wherein the first factor data 22 (hereinafter referred to as factor data 22 - 1 ), the second factor data 22 (hereinafter referred to as factor data 22 - 2 ) and the result data 24 are associated with each other.

[0128] The calculation unit 2040 divides each of the numerical range of the factor data 22-1 and the numerical range of the factor data 22-2 into the number of divisions indicated by the division number information 30. Note that the division number information 30 may indicate a common number of divisions between the factor data 22-1 and the factor data 22-2, or may indicate a different number of divisions for each of the factor data 22-1 and the factor data 22-2. For example, it is assumed that the number of divisions of the factor data 22-1 is 4 and the number of divisions of the factor data 22-2 is 5. In this case, the calculation unit 2040 divides the numerical range of the factor data 22-1 into four equal parts and divides the numerical range of the factor data 22-2 into five equal parts. Specifically, the calculation unit 2040 determines three boundary values ​​for the factor data 22-1 and four boundary values ​​for the factor data 22-2.

[0129] For each of the plurality of pairs of the boundary value of the factor data 22-1 and the boundary value of the factor data 22-2, the computing unit 2040 divides the object data set 10 into two parts at the boundary defined by the pair, and generates a first sample and a second sample. Then, as described above, the computing unit 2040 calculates the test statistic for the generated first sample and the second sample.

[0130] Fig.12 2 is a schematic diagram showing a first sample and a second sample in an analysis device 2000 according to a modified example. In this example, the first sample and the second sample are generated at a boundary defined by a pair of a boundary value b2 of the factor data 22-1 and a boundary value c2 of the factor data 22-2. Specifically, the calculation unit 2040 generates a set of object data 20 that satisfies the condition "factor data 22-1 is b2 or less and factor data 22-2 is c2 or less" as the first sample 12, and generates a set of object data 20 that satisfies the condition "factor data 22-1 is greater than b2, or factor data 22-2 is greater than c2" as the second sample 14.

[0131] However, as described above, as a method of dividing the object data set 10 into two samples by using a pair of two boundary values, it is possible to adopt Figure 1 A method other than the method shown (hereinafter referred to as the first method). Fig.13 FIG. 1 is a schematic diagram showing another method of dividing the object data set 10 into two samples using a pair of two boundary values. Fig.13As shown, in addition to the above methods, there is a second method of dividing into a part where "factor data 22-1 is equal to or less than the boundary value, and factor data 22-2 is equal to or greater than the boundary value" and a part where "factor data 22-1 is greater than the boundary value, or factor data 22-2 is less than the boundary value"; a third method of dividing into a part where "factor data 22-1 is equal to or greater than the boundary value, and factor data 22-2 is equal to or less than the boundary value" and a part where "factor data 22-1 is less than the boundary value, or factor data 22-2 is greater than the boundary value"; a fourth method of dividing into a part where "factor data 22-1 is equal to or greater than the boundary value, and factor data 22-2 is equal to or greater than the boundary value" and a part where "factor data 22-1 is less than the boundary value, or factor data 22-2 is less than the boundary value"; and so on. Which of these methods is used to divide the object data set 10 can be pre-defined or selected by the user.

[0132] Fig.14 is a schematic diagram showing information to be output by an analysis device 2000 according to a modification. Fig.14 In the screen 140 shown, Fig.11 The information shown in screen 130 applies to the case where two factors are used.

[0133] The screen 140 includes display areas 141 to 145. The display area 141 includes a time chart 100 showing time changes of each of the factor data 22-1 and the factor data 22-2. The display area 142 includes a histogram 90-1 of the factor data 22-1 and a histogram 90-2 of the factor data 22-2.

[0134] Similar to the display area 133, the display area 143 includes an input interface and a button, the input interface is used to specify the test method and the test granularity, and the button is used to instruct the test. The display area 143 also includes an input interface 146 for specifying the method of dividing into the first sample and the second sample. The input interface 146 includes four areas composed of four partial rectangles and is configured to enable the user to specify one of the four areas. A color is given to the specified area.

[0135] The input interface 146 simulates a plane consisting of the value ranges of the two factor data 22, which is Fig.12 and Fig.13 When the user specifies one of the four regions, the object data set 10 is divided into two samples so that the plane formed by the value ranges of the two factor data 22 is divided into the specified region and the region other than the specified region. Fig.14In the example of , the upper left area is specified. Therefore, the object data set 10 is divided into two samples, that is, the part where "factor data 22-1 is equal to or less than the boundary value, and factor data 22-2 is equal to or less than the boundary value" and the part other than that.

[0136] The display area 144 shows a test statistic graph 150. For each of a plurality of pairs of boundary values ​​of the factor data 22-1 and boundary values ​​of the factor data 22-2, the calculation unit 2040 calculates a test statistic with respect to two samples obtained by the pair and generates statistical information 40 (here, the test statistic graph 150) representing a plurality of test statistics. Each cell of the test statistic graph 150 represents a test statistic obtained for a pair of boundary values ​​associated with the cell. For example, in Fig.14 In the example of , each cell represents the magnitude of the test statistic obtained for the pair of boundary values ​​represented by the lower right end of the cell.

[0137] In the test statistic graph 150, the size of the test statistic is represented by color, pattern, etc. For example, as the test statistic increases, the color becomes darker, or the pattern becomes thicker. Fig.14 In the example of , a larger dot pattern is given to a cell with a large test statistic. In addition, a border 152 is given to a cell with a test statistic equal to or greater than a threshold value th.

[0138] The user can specify a pair of desired boundary values. Specifically, the user specifies a unit associated with the pair of desired boundary values. In response to the specification, the generation unit 2060 generates information related to the improvement of the defect rate of the sample pair to be generated by the specified unit and the specified sample division method.

[0139] exist Fig.14 In the example, a pair of cells representing boundary values ​​"boundary value of component A = b2, boundary value of component B = c2" is specified. In addition, a method of dividing the plane of the numerical range into an upper left portion and a portion other than that is specified. In view of this, the plane of the numerical range is divided into samples of "value of component A ≤ b2 and value of component B ≤ c2" and samples of "value of component A > b2 or value of component B > c2". Note that the display scheme of histograms 90-1 and 90-2 changes in response to the specification. Specifically, as described in exemplary embodiment 1, between levels whose values ​​are equal to or less than the specified boundary value and levels other than that, colors or patterns are set to be different from each other.

[0140] The generation unit 2060 displays information indicating the simulation of the improvement of the defect rate for the above division in the display area 145. The display area 145 indicates the comparison of the number of defective products, the number of qualified products, and the comparison of the defect rate with respect to the object data 20 satisfying the condition "setting component A to be equal to or less than b2, and setting component B to be equal to or less than c2" and the defect rate in all the object data 20. Through the above comparison, the user can easily recognize to what extent the defect rate can be improved by manufacturing the product under the manufacturing condition of "setting component A to be equal to or less than b2, and setting component B to be equal to or less than c2".

[0141] Although the exemplary embodiments of the present invention have been described with reference to the accompanying drawings, the above exemplary embodiments are merely examples of the present invention, and various configurations in combination with or in addition to the above exemplary embodiments may be adopted.

[0142] A part or all of the above-described exemplary embodiments may be described as the following supplementary notes but are not limited thereto.

[0143] 1. An analysis device comprising:

[0144] an acquisition unit that acquires a set of object data and division number information, each object data indicating a pair of factor data and result data, the factor data being a value associated with a factor, the result data being a value associated with a result, and the division number information indicating the number of numerical ranges for dividing the factor data;

[0145] a calculation unit that calculates a test statistic for each of a plurality of boundary values, the plurality of boundary values ​​being used to divide a numerical range into quantities, the test statistic representing a difference associated with result data of two samples to be obtained by dividing object data included in a set of object data into two parts at the boundary value; and

[0146] A generating unit generates test statistic information indicating a plurality of test statistics calculated for each boundary value.

[0147] 2. The analysis device according to Supplementary Note 1, wherein:

[0148] The resulting data indicates the presence or absence of defects, and

[0149] The calculation unit calculates a defect rate of each of the two samples for each boundary value, and calculates a test statistic representing a difference between the calculated two error rates.

[0150] 3. The analysis device according to Supplementary Note 2, wherein:

[0151] The factor data indicates a value of an index representing a manufacturing condition of a product, and

[0152] The result data indicates whether the product is a defective product.

[0153] 4. The analysis device according to any one of Supplementary Notes 1 to 3, wherein:

[0154] The test statistic information is a graph indicating a combination of a boundary value and a test statistic.

[0155] 5. The analysis device according to Supplementary Note 4, wherein:

[0156] The generation unit includes an indication in the graph representing a threshold value of the test statistic used to represent the significance level.

[0157] 6. The analysis device according to Supplementary Note 4 or 5, wherein:

[0158] Generate unit together with Figure 1 It outputs a histogram showing the distribution of the factor data.

[0159] 7. The analysis device according to Supplementary Note 6, wherein:

[0160] Generate Unit

[0161] accepts input that specifies bounds for the graph, and

[0162] In response to accepting input, in the histogram, the display scheme of data in the level associated with each boundary value equal to or less than the specified boundary value and the display scheme of data in the level associated with each boundary value greater than the specified value are different from each other.

[0163] 8. The analysis device according to any one of Supplementary Notes 1 to 7, wherein:

[0164] The object data indicates a collection of the first factor data, the second factor data, and the result data,

[0165] The calculation unit calculates, for each set of multiple first boundary values ​​and multiple second boundary values, a test statistic of two samples to be obtained by dividing the object data included in the object data set into two parts based on the set of boundary values, the first boundary value is used to divide the numerical range of the first factor data into quantities, and the second boundary value is used to divide the numerical range of the second factor data into quantities.

[0166] 9. A control method to be executed by a computer, comprising:

[0167] an acquisition step of acquiring a set of object data and division quantity information, each object data indicating a pair of factor data and result data, the factor data being a value associated with a factor, the result data being a value associated with a result, and the division quantity information indicating the number of numerical ranges for dividing the factor data;

[0168] a calculating step of calculating a test statistic for each of a plurality of boundary values, the plurality of boundary values ​​being used to divide a numerical range into quantities, the test statistic representing a difference associated with result data of two samples to be obtained by dividing object data included in a set of object data into two parts at the boundary value; and

[0169] A generating step generates test statistic information indicating a plurality of test statistics calculated for each boundary value.

[0170] 10. The control method according to Supplementary Note 9, wherein:

[0171] The resulting data indicates the presence or absence of defects.

[0172] The control method further comprises,

[0173] In the calculation step, the defect rate for each of the two samples is calculated for each boundary value, and a test statistic representing the difference between the two calculated error rates is calculated.

[0174] 11. The control method according to Supplementary Note 10, wherein:

[0175] The factor data indicates a value of an index representing a manufacturing condition of a product, and

[0176] The result data indicates whether the product is a defective product.

[0177] 12. The control method according to any one of Supplementary Notes 9 to 11, wherein:

[0178] The test statistic information is a graph indicating a combination of a boundary value and a test statistic.

[0179] 13. The control method according to Supplementary Note 12 further comprises:

[0180] In the generating step, an indication representing a threshold value representing a significance level of the test statistic is included in the plot.

[0181] 14. The control method according to Supplementary Note 12 or 13 further comprises:

[0182] In the Generate step, together with Figure 1 It outputs a histogram showing the distribution of the factor data.

[0183] 15. The control method according to Supplementary Note 14 further comprises:

[0184] In the Generate step,

[0185] accepts input that specifies bounds for the graph, and

[0186] In response to accepting input, in the histogram, the display scheme of data in the level associated with each boundary value equal to or less than the specified boundary value and the display scheme of data in the level associated with each boundary value greater than the specified boundary value are different from each other.

[0187] 16. The control method according to any one of Supplementary Notes 9 to 15, wherein:

[0188] The object data indicates a collection of the first factor data, the second factor data, and the result data,

[0189] The control method further comprises:

[0190] In the calculation step, for each set of multiple first boundary values ​​and multiple second boundary values, a test statistic of two samples to be obtained by dividing the object data included in the object data set into two parts based on the set of boundary values ​​is calculated, the first boundary value is used to divide the numerical range of the first factor data into quantities, and the second boundary value is used to divide the numerical range of the second factor data into quantities.

[0191] 17. A program causing a computer to execute the control method according to any one of Supplementary Notes 9 to 16.

[0192] This application is based upon and claims the benefit of priority from Japanese Patent Application No. 2019-199613, filed on November 1, 2019, the disclosure of which is incorporated herein in its entirety by reference.

[0193] Reference numerals list

[0194] 10 Object Data Collection

[0195] 12 First Sample

[0196] 14 Second Sample

[0197] 20 Object data

[0198] 22 Factor Data

[0199] 24 Results Data

[0200] 30 Division quantity information

[0201] 40 Test statistic information

[0202] 50 Table

[0203] 52 Boundary value

[0204] 54 Test Statistic

[0205] 60 User Terminals

[0206] 70 Test Statistic Plot

[0207] 72 articles

[0208] 74 Threshold indication

[0209] 76 Instructions

[0210] 80 screens

[0211] 82,84,86 Input area

[0212] 88 Button

[0213] 90 Histogram

[0214] 100 Timing Diagram

[0215] 102 lines

[0216] 110 lines

[0217] 130 screens

[0218] 131,132,133,134,135,136,137 Display area

[0219] 140 screens

[0220] 141,142,143,144,145 Display area

[0221] 146 Input Interface

[0222] 150 Test Statistic Plot

[0223] 152 Frames

[0224] 1000 Computer

[0225] 1020 Bus

[0226] 1040 processor

[0227] 1060 Memory

[0228] 1080 Storage Devices

[0229] 1100 Input / Output Interface

[0230] 1120 Network Interface

[0231] 2000 Analytical Devices

[0232] 2020 Acquisition Unit

[0233] 2040 Computing Unit

[0234] 2060 Generation Unit

Claims

1. An analysis device comprising: an acquisition unit, the acquisition unit acquiring a set of object data and division number information, each object data indicating a pair of factor data and result data, the factor data being a value associated with a factor, the result data being a value associated with a result, and the division number information indicating a number of numerical ranges for dividing the factor data; a calculation unit that calculates a test statistic for each of a plurality of boundary values ​​for dividing the numerical range into the quantities, the test statistic representing a difference associated with the result data of two samples to be obtained by dividing the object data included in the set of object data into two parts at the boundary value; as well as A generating unit, wherein the generating unit generates test statistic information, wherein the test statistic information indicates a plurality of test statistics calculated for each boundary value, wherein: The result data indicates the presence or absence of defects, and The calculation unit calculates a defect rate of each of two samples for each boundary value, and calculates a test statistic representing a difference between the calculated two defect rates.

2. The analysis device according to claim 1, wherein The factor data indicates a value of an index representing a manufacturing condition of a product, and The result data indicates whether the product is a defective product.

3. The analysis device according to claim 1 or 2, wherein: The test statistic information is a graph indicating a combination of the boundary value and the test statistic.

4. The analysis device according to claim 3, wherein: The generating unit includes in the graph an indication representing a threshold value of the test statistic representing a significance level.

5. The analysis device according to claim 3, wherein: The generation unit outputs a histogram representing the distribution of the factor data together with the graph.

6. A control method to be executed by a computer, comprising: Acquire a set of object data and division quantity information, each object data indicating a pair of factor data and result data, the factor data being a value associated with a factor, the result data being a value associated with a result, and the division quantity information indicating the number of numerical ranges for dividing the factor data; calculating a test statistic for each of a plurality of boundary values, the plurality of boundary values ​​being used to divide the numerical range into the quantities, the test statistic representing a difference associated with the result data of two samples to be obtained by dividing the object data included in the set of object data into two parts at the boundary value; as well as Generate test statistic information, the test statistic information indicating a plurality of test statistics calculated for each boundary value, wherein, The result data indicates the presence or absence of defects, and The defect rate for each of the two samples is calculated for each boundary value, and a test statistic representing the difference between the two calculated defect rates is calculated.

Images