JP5778530B2 - Question item selection method - Google Patents

Description

The present invention relates to a question item selection method, in particular, relates to the question item selection method for survey data analysis by computer system factor analysis program is incorporated.

  For example, questionnaire surveys are frequently performed in order to understand market trends and to conduct various types of analysis and research. In the questionnaire survey, questionnaire items related to the contents to be surveyed are examined and selected in advance, and data is collected by having a questionnaire target person answer the selected questionnaire items. In order to effectively conduct questionnaire surveys, select appropriate question items related to the survey contents as questionnaire items so that data can be collected efficiently, and the number of questionnaire items must be increased excessively. It is necessary not to create a burden on the survey subjects.

  On the other hand, multivariate analysis is known as a statistical processing method for analyzing a large number of data collected by a questionnaire survey. According to multivariate analysis, data analysis is performed efficiently and quickly by using various analysis programs incorporated in a computer system. For example, a predetermined number of questionnaire subjects are determined based on common factors (factors). It can be grouped into groups. A factor analysis program is known as a program for performing multivariate analysis (see, for example, Patent Document 1). Factor analysis is a statistical method that uses a plurality of variables simultaneously to clarify the structure of the relationship between these variables, and extracts and analyzes factors common to the measurement values of a plurality of data.

JP 2004-164078 A

  However, conventional questionnaire data for analysis by the factor analysis program is obtained as an answer to questionnaire items that are artificially selected to meet the survey objectives of the questionnaire, and is also analyzed based on the questionnaire data. The characteristics of the common factors obtained as a result of the survey are not absolute, and are considered optimal in view of the contents of questionnaire items highly correlated with the factors and the factor loadings for each questionnaire item. It is designed to be read and read appropriately. Therefore, when questionnaires were conducted for the same survey purpose at different times and places, even if the questionnaires were conducted with the same questionnaire items, they were obtained by conducting an analysis each time they were selected artificially. It becomes difficult to extract factors related to the same characteristics from multiple common factors, and the questionnaire data from questionnaires conducted for the same research purpose at different times and places are compared as questionnaire data for evaluating the same characteristics. It becomes difficult to use.

  In addition, when using a factor analysis program to classify survey targets with high accuracy or to find hidden factors, it is necessary to increase the number of questionnaire items. In other words, it is necessary for accurate classification to conduct a questionnaire including items whose relation to the purpose of the questionnaire is unknown, analyze the questionnaire data obtained as a result, extract factors, and classify the survey targets It becomes.

  On the other hand, as the number of questionnaire items increases, it takes time for the survey subjects, making it difficult to obtain cooperation. Will have an impact.

The present invention provides a question item selection method for questionnaire data analysis that makes it possible to use questionnaire data from a questionnaire conducted for the same research purpose at different times and places as questionnaire data for evaluating the same characteristics. For the purpose.

It is another object of the present invention to provide a question item selection method for questionnaire data analysis that enables at least survey items to be classified (grouped) with high accuracy.

  The present invention is a question item selection method for questionnaire data analysis by a computer system in which a factor analysis program is incorporated, and by collecting a large number of questionnaire data by a plurality of question items and performing factor analysis, Factor extraction processing step for extracting factors, and questionnaire items that have a factor loading less than a predetermined absolute value are rounded down as question items that are weakly correlated with each factor extracted in the factor extraction processing step. A question item truncation processing step that selects only question items that have a strong correlation with each factor, and a question item that is highly correlated with each factor selected in the question item truncation processing step is positive only with a specific factor. An effective question selection processing step that selects a question item having a strong correlation as an effective question, and is selected in the effective question selection processing step. Including effective question questionnaire, by providing a question item selecting method questionnaire regarding the specific factor is created, is obtained by achieving the above object.

According to the question item selection method for questionnaire data analysis of the present invention, it is possible to use questionnaire data obtained by questionnaires conducted for the same investigation purpose at different times and places as questionnaire data for evaluating the same characteristics. be able to. Moreover, even if there are few survey items, it is possible to classify (group) the survey objects with high accuracy.

1 is a system configuration diagram of a computer system for implementing a question item selection method according to a preferred embodiment of the present invention. 5 is a flowchart illustrating a question item selection method according to a preferred embodiment of the present invention. It is a chart which illustrates the question table in which data contents, such as a question sentence, were displayed as a questionnaire item. It is a chart which illustrates a question table after extracting an effective question. It is a graph which illustrates the data of the answer result by the some questionnaire respondent regarding the extracted questionnaire item. It is a chart which illustrates the analysis result by a factor extraction processing step. It is a chart which illustrates the analysis result by a factor extraction processing step more concretely. It is a flowchart explaining a question item truncation processing step. It is a chart which illustrates the analysis result after rewriting which judged whether the absolute value of factor load amount exceeded a threshold. It is a graph which illustrates more concretely the analysis result after rewriting which judged whether the absolute value of factor load amount exceeded a threshold. It is a flowchart explaining an effective question selection process step. It is a chart which illustrates concretely an analysis result after an effective question selection processing step.

  The question item selection method according to a preferred embodiment of the present invention is performed using a computer system 10 including a computer 11 having a database 12, as shown in FIG. For the purpose of obtaining appropriate data to support future dietary habits, behavioral styles, etc., for example, to be able to select effective questionnaire items related to health status and its management, etc. It was adopted to do.

  The question item selection method of the present embodiment is a question item selection method for questionnaire data analysis by the computer system 10 in which the factor analysis program is incorporated, and as shown in FIG. By collecting a large number of questionnaire data and performing factor analysis, factor extraction processing step S1 for extracting a common factor, and question items that are weakly correlated with each factor extracted in factor extraction processing step S1, Question items are selected at the question item truncation processing step S2 and the question item truncation processing step S2 to select only the question items having a strong correlation with each factor by truncating the questionnaire items having factor loadings below a predetermined absolute value. Of the question items that have a strong correlation with each factor, select the question items that have a strong positive correlation only with specific factors as valid questions. That effective question selection process includes a step S3, an effective question selected in effective question selection process step S3 is included in the questionnaire item, creating a combination of a questionnaire i.e. more questions about that particular factor.

  The questionnaire survey is performed by creating a questionnaire form by printing a combination of a plurality of questions including the valid question selected in the questionnaire thus created, that is, the valid question selection processing step S3, or by using a personal computer, a mobile phone device, etc. An interviewer who asks for an answer or asks for an answer is displayed on the screen of the electronic device and the screen of the electronic device as shown above on which the questionnaire is printed or the questionnaire is displayed. Or asking questions and collecting answers.

  In the present embodiment, the computer 11 constituting the computer system 10 shown in FIG. 1 that implements the question item selection method and the questionnaire form creation method is known as a computer including a database 12, and the database 12 includes a CPU, ROM, RAM, I / F, storage means, input means, display means, output means and the like are provided.

  The CPU of the computer 11 controls the overall operation of the database 12 while using the RAM as a work area in accordance with a control program incorporated in the ROM. In addition, since various data analysis programs and other computer programs are incorporated in a program storage unit (not shown) of the hard disk, the CPU can analyze questionnaire data by factor analysis, Various other processes such as display and printing are performed. The storage means of the database 12 stores, for example, questionnaire data and other data relating to various health conditions and management thereof in the data storage unit 13 constituted by a hard disk, and reads the stored data to analyze data. Used for other processing. Further, the computer 11 is connected to a display unit 14 formed of a display device, a printing unit 15 formed of a printer, an input unit 16 formed of a keyboard and a mouse, and the like.

  In the present embodiment, the CPU of the computer 11 incorporates a factor analysis program as a data analysis program, so that it can be appropriately extracted from various questionnaire data stored in the data storage unit 13, for example. Factor analysis can be performed on the basis of the obtained data. As the factor analysis program, preferably known “SPSS” (manufactured by SPS Corporation) can be used as statistical analysis software.

  In the present embodiment, as shown in FIG. 2, for example, after “SPSS” is started, a process for extracting questionnaire items suitable for performing factor analysis is performed. That is, a plurality of questionnaire items that are considered to be appropriate for classifying survey target persons (questionnaire respondents) from a large number of questionnaire items in preliminary questionnaires stored in the data storage unit 20 together with the questionnaire results. To extract. Such a process of extracting questionnaire items is easily performed by selecting an appropriate questionnaire item from a question table in which data contents such as a question sentence are displayed together with an address as shown in FIG. 3, for example. be able to. Note that the column of the valid flag in the question table is initially s (1) is 1, but after the questionnaire item extraction processing, as shown in FIG. 4, “1” (valid) or “0”. It will take one of the values (invalid).

  In the factor extraction processing step S1 of the question item selection method of the present embodiment, based on the answer result data as shown in FIG. 5 by a plurality of questionnaire respondents regarding each questionnaire item in the preliminary questionnaire described above. By performing analysis, a process for extracting factors common to each questionnaire item is performed. Such factor analysis can be easily performed using the above-described “SPSS” function. It is also preferable to select the principal factor method as the analysis method and maximize the variance of the factor loading by varimax rotation. The analysis result of the questionnaire data by the factor extraction processing step S1 is shown in FIG.

  Here, the values of a large number of questionnaire data by a plurality of questionnaire respondents for a plurality of questionnaire items as shown in FIG. 5 for performing factor analysis are, for example, five integers from 0 to 4, taking an integer. The questionnaire data is preferably obtained as an order variable. Since the values of a large number of questionnaire data for performing factor analysis are integers obtained as order variables from the five answer choices, an accurate analysis can be performed. In addition, as a result, for example, as specifically shown in FIG. 7, it is possible to easily obtain the analysis result of the questionnaire data in which the factor load is expressed as a real number.

  In the question item truncation processing step S2 of the question item selection method of the present embodiment, the factor load amount lower than a predetermined absolute value as a question item weakly correlated with each factor extracted in the factor extraction step S1 The questionnaire items that become are truncated, and only the question items that are highly correlated with each factor are selected. In the question item truncation processing step S2, according to the flowchart shown in FIG. 8, first, the threshold RTH in which the absolute value of each factor load (R (i, j)) in the chart shown in FIG. 6 is a predetermined absolute value. It is determined whether it becomes less than. If the absolute value of the factor loading (R (i, j)) is less than the threshold value RTH, replace (R (i, j)) with “0” and factor loading (R (i, j)) (R (i, j)) is replaced with (r (i, j)) when the absolute value of is equal to or greater than the threshold value RTH. That is, in FIG. 9 in which (R (i, j)) in FIG. 6 is replaced with (r (i, j)), the absolute value of (R (i, j)) is less than the threshold value RTH (R (i, j)) is replaced and takes a value of “0”. In FIG. 9 showing (R (i, j)) after replacement, that is, (r (i, j)), all the values of (r (i, j)) in the row direction are “ The question items that are “0” are rounded down as question items that are weakly correlated with each factor, while the question items whose row direction (r (i, j)) values are not all “0” It is stored together with the address as a valid question candidate having a strong correlation with any of the factors.

  More specifically, after rewriting after determining whether or not the absolute value of each factor load amount in the chart of FIG. 7 is less than “0.4” by setting the predetermined threshold value RTH to “0.4”. In the chart of FIG. 10 showing the analysis results, the Q2 and Q5 question items in which the values of the factor loadings in the row direction are all “0” are rounded down as question items that are weakly correlated with both factors. Thus, the question items Q1, Q3, Q4, and Q6 are selected as effective question candidates having a strong correlation with at least one factor, and stored together with the address.

  In the valid question selection processing step S3 of the question item selection method of the present embodiment, among the question items (valid question candidates) having a strong correlation with each factor selected in the question item truncation processing step S2, Only question items with strong positive correlation are selected as valid questions. That is, in the valid question selection processing step S3, according to the flowchart shown in FIG. 11, the values of (r (i, j)) in the row direction are all “0” in the chart of FIG. Of the question items (valid question candidates) that are strongly correlated with any of the factors, (r (i, j)) for factor k is not “0” and all other than factor k The question item with (r (i, j)) of “0” or minus for the factor is selected as an effective question regarding the factor k.

  More specifically, in the chart of FIG. 10 showing the analysis result after rewriting after determining the predetermined threshold value RTH as “0.4”, the factor load amount for factor 4 is not “0”. As shown in FIG. 12, Q7 question items that are “0.5” and the factor loadings for all factors 1 to 3 other than factor 4 are “0” or minus, It becomes possible to select an effective question having a so-called one-factor structure regarding the factor 4.

  In the present embodiment, the question item Q7 having the one-factor structure obtained as described above is selected as an effective question, and the selected effective question is included in the questionnaire item, for example, regarding the specific factor 4 An effective questionnaire will be created for the survey.

  The created questionnaire is processed into printing data by using well-known insertion printing software attached to commercially available word processing software, which is incorporated in the program storage unit of the CPU. Sent and printed. In this way, a questionnaire form including the selected valid question as a questionnaire item is created.

  Then, according to the question item selection method and the questionnaire form creation method of the present embodiment having the above-described configuration, a small number of questionnaire items selected from a large number of questionnaire items by the preliminary questionnaire are used as questionnaire items for performing the same classification. Can be used.

  That is, according to the question item selection method of the present embodiment, a factor extraction processing step S1 for collecting a large number of questionnaire data and performing a factor analysis to extract a common factor and a predetermined absolute value determined in advance. Question item truncation processing step S2 in which questionnaire items that are less than factor loading are cut off and only the question items that are highly correlated with each factor are selected, and question items that are positively correlated only with specific factors are valid questions A valid question selection process step S3 to be selected. The valid question having the one-factor structure selected in the valid question selection process step S3 is included in the questionnaire item, and preferably selected in the valid question selection process step S3. Surveys are created using only valid questions with a one-factor structure as questionnaire items. It becomes possible to perform the over door.

  By printing the questionnaire created in this way with the printing unit 15 as described above, it is possible to create a questionnaire form with a small number of question items, that is, a small number of pages, and to reduce the burden on the respondents of the questionnaire. As a result, it becomes possible to improve the collection rate of responses to questionnaires.

  The present invention is not limited to the above-described embodiment, and various modifications can be made. For example, a large number of questionnaire data based on a plurality of questionnaire items for performing factor analysis does not necessarily need to be questionnaire data obtained as an order variable from five answer options. As the factor analysis program, other known factor analysis programs other than “SPSS” may be used.

  In addition to the valid question having the one-factor structure selected in the valid question selection processing step S3, it is optional to create a questionnaire by adding other questions as questionnaire items. The questionnaire items added in this way can be used as candidates for new valid questions, and the contents of the questionnaire items can be changed according to changes in the times and the environment.

DESCRIPTION OF SYMBOLS 10 Computer system 11 Computer 12 Database 13 Data storage part 14 Display part 15 Printing part 16 Input part

Claims (2)

A question item selection method for questionnaire data analysis by a computer system incorporating a factor analysis program,
Factor extraction processing step to extract common factors by collecting a large number of questionnaire data for multiple questionnaire items and performing factor analysis,
As question items that are weakly correlated with each factor extracted in the factor extraction processing step, question items that have a factor loading less than a predetermined absolute value are rounded down, and question items that are highly correlated with each factor Question item truncation processing step to select only,
An effective question selection processing step of selecting, as an effective question, a question item having a strong positive correlation only with a specific factor among the question items having a strong correlation with each factor selected in the question item truncation processing step,
A question item selection method in which a valid question selected in the valid question selection processing step is included in a questionnaire item, and a questionnaire related to the specific factor is created.   The question item selection method according to claim 1, wherein a large number of questionnaire data based on the plurality of questionnaire items for performing factor analysis is questionnaire data obtained as an order variable from five answer options.

Images