SOCIAL RESEARCH DATABASE on QUESTIONNAIRES

Using the Analysis Page

1. Top Page

Fig. 1 shows the SRDQ top page. The box at the bottom left of this page shows the "social research data that can be analyzed" using the browser. There are " Overview" of the research and "Analysis" pages for each of the researches. Here, we will show how to use the "Analysis" page.

Fig. 1 The SRDQ Top Page
Fig. 1 The SRDQ Top Page

2. Dataset Analysis

From the "Social Research Data That Can Be Analyzed" box at the bottom left of the top page (Fig. 1), select the research that you want to analyze. In this explanation, we will use "Japan Survey on Information Society 2001 (JIS2001)" as an example. Click the blue "Analysis" button under the "Japan Survey on Information Society 2001 (JIS2001)" (orange text). A security warning is displayed; simply disregard the warning and click "Continue". The page shown in Fig. 2 is displayed.

Fig. 2 Dataset Analysis Page
Fig. 2 Dataset Analysis Page

The title of the research data to be analyzed ("Japan Survey on Information Society 2001 (JIS2001)") is displayed in blue under the page title of "Dataset Analysis". The next line contains links to “Overview of the Research” (see Fig. 2 for information about “Using the Search Page”) and to “List of the Questions” (see Fig. 4 for information about “Using the Search Page”) for the JIS 2001 research. The "Analysis Menu" below these links shows the analysis methods that can be used with SRDQ. The SRDQ analysis functions use WebApp from SPSS Japan Inc. With WebApp, all of the functions included in the SPSS Base and options can be used. However, customization is needed to utilize all of the available functions, so at present, SRDQ is limited to the seven statistical methods shown in the analysis menu - Frequencies, Summarize, Cross tabs, Web Cubes, T-Test, One-way ANOVA, and Linear Regression, and to certain options for each type of analysis.

Now, let's look at each of these analytical methods.

3. Frequencies

Select "Frequencies" from the "Analysis Menu" shown in Fig. 2. The page shown in Fig. 3 is displayed.

Fig. 3 Frequencies
Fig. 3 Frequencies

The data box on the left side of the page shows a list of the variable names used in the "Japan Survey on Information Society 2001 (JIS2001)". The list of variables continues beyond what is displayed in the data box, and any of the variables can be selected. Fig. 4 shows the output when "Q3_1: Cellular phone" is selected from the list.

Fig. 4 Frequencies Output
Fig. 4 Frequencies Output

The "Statistics" box at the top center of the page shows the number of effective replies(Valid) and the number of missing data items(Missing). The next box shows the frequency distribution table. The box at the bottom of the page shows a bar graph. A list of the variable names is also shown on the left side of this page; selecting a different variable displays the frequency distribution table output for that variable.

4. Descriptives

You can refer to the basic statistics on variables included in the analyzable dataset. The procedure for referencing the basic statistics on the variable “Q2: Age” included in the dataset in the Japan Survey on Information Society (JIS2001) is described here.

Procedure
  1. Click on Descriptives on the Analysis Menu window. A new window as shown in Figure 5 will appear.
  2. Select the name of a variable (Q2: Age) on which you want to refer to the basic statistics from the list of variables on the left side of Select and click on the >> button. You can refer to the basic statistics on two or more variables at a time. If you select the wrong variable, click on the << button and change it.
  3. In Statistics, select the statistics to which you want to refer. The statistics to which you can refer are Mean, Median, Mode, and Sum in Central Tendency; Std. deviation, Variance, Range, Minimum, and Maximum in Dispersion; Percentile Values in Quartiles; and Skewness and Kurtosis in Distribution. By clicking on Select all items or Release all items, you can select all items or release them at a time. In the present example, you select Mean, Median, Mode, Std. deviation, Range, Minimum, Maximum, Skewness, and Kurtosis.
  4. After completing tasks i to iii, click on the Descriptives button and the results shown in Figure 6 will appear. In addition to the statistics you have selected, Valid and Missing values will be displayed. In Skewness and Kurtosis, Std. Error will also be indicated.
Fig. 5 Select and Statistics
Fig. 5 Select and Statistics
Fig. 6 Statistics
Fig. 6 Statistics

5. Summarize

Select " Summarize " from the “Analysis Menu” shown in Fig. 2. The page shown in Fig. 7 is displayed.

Clicking the down arrow to the right of "Select a Measure" for item 1 displays a drop-down menu with a list of the variables that can be selected. From this menu, select the variable for which you want to display the group sum, averages and/or number of cases. In section 2 "Statistics", select the necessary statistics -sum, averages and/or number of cases in each group. Any or all of the items can be selected. In section 3 "Group Selection", select the variables to be specified for the group. This list has been designed so that not many category variables, what is called qualitative variables can be selected. Up to three variables can be selected. After making selections for items 1, 2 and 3, click the "go" button in item 4.

Fig. 7 Summarize
Fig. 7 Summarize

Fig. 8 shows the results displayed when "Q3_1: Cellular phone" is selected for the measurement, "Sum", "Average" and "Number of Cases in each group" are selected for the type of statistics, and "Age 3 categories" is selected for the group selection.

Fig. 8 Summarize Output
Fig. 8 Summarize Output

The results are output in the "Report" box on the right side. The same variable specification box shown in Fig. 7 is also shown as is on this page. Additional case summary analyses can be made by specifying different variables in the variable specification box.

6. Crosstabs

Select "Crosstabs" from the "Analysis Menu" shown in Fig. 2. The page shown in Fig. 9 is displayed.

The variables are displayed in the box on the left. Scroll down through the list and select the variables to be analyzed. Select variables for use as row variables and column variables to generate a standard cross table. To generate a three-layer cross table, also select layer variables. Multiple variables can be selected for the row variables, column variables and layer variables, with the selected variables displayed from top to bottom in the corresponding boxes to the right. To change the order position of a variable in each box, select the corresponding variable and then click the "up" or "down" button located just to the right of the variable list box.

The "Option" settings section on the right side of the page can be used to set the values for "Chi-Square" and "Phi and Cramer’s V". If the "Row" percentage item is selected for the "Cell Display Setting", each cell shows a row percentage; if "Column" is selected, each cell shows a column percentage. The sorting order for the rows can also be set to display the rows in ascending or descending order. The default settings are chi-square, row percentage, and ascending order.

Fig. 9 Crosstabs
Fig. 9 Crosstabs

Fig. 10 shows the results when the row variable is set to "Q1: Gender", the column variable to "Q3_8: PC", and the option settings are set to chi-square, row percentage, and ascending order.

Fig. 10 Crosstabs Output
Fig. 10 Crosstabs Output

Click "Back to Cross Tabulations" at the top of the page to return to the page shown in Fig. 9. The "Case Processing Summary" shown below the back link displays the number of valid cases and the ratio of valid cases to the total number of cases. The section in the middle shows the resulting cross tabulation table. The section at the bottom shows the results of χ2 test.

The output formats for this and other types of analyses are nearly the same as the formats available in using SPSS directly from a PC. However, the icons shown at the top left of each section may be unfamiliar. Moving the mouse cursor over the icon at the top (spss csv) displays the icon function, "Download data as CSV file" in this case. Click this icon to save the data shown in the corresponding section to your computer as a CSV file. CSV files can be read by most spreadsheet programs such as MS Excel. The second icon from the top (new window) is the "Open in new window" icon; clicking this icon displays the data shown in the corresponding section in a new window. The third icon from the top (flip table) is the "Flip table" icon, and is used to flip the row and column variables and display the results. The last icon (help) is the "Help" icon; clicking this icon displays a simple explanation of the corresponding section in English.

7. Web Cubes

Select "Web Cubes" from the “Analysis Menu” shown in Fig. 2. The page shown in Fig. 11 is displayed.

This page is divided into four sections, "Test Measures", "Grouping Dimensions", "Statistics" and "Display Options". Only one of the "Test Measures" can be selected. Most of the sections are the same as those already described for the other analytical methods, but the "Display Options" section is unique to the Web cubes method.

Either "Measurs" or "Statistic" can be selected for comparison within data cells, "Drilldown Dimensions" "or "Nest Dimensions" can be selected for comparison within rows, and "Above table" or "Below table" can be selected when a slice dimension is used. Title and Caption can also be entered in both English and Japanese.

Fig. 11 Web Cubes
Fig. 11 Web Cubes

Fig. 12 shows the page output when "TQ9DAY: Looking at websites a day (minutes)" is selected as the measurement subject variable, "Age 3 categories" and "Citysize 3 categories" are selected as the group variables, and "Average" is selected as the statistic. When using the PC version of SPSS, nearly the same results can be obtained by using OLAP cubes with the pivot function.

Fig. 12 Web Cubes Output
Fig. 12 Web Cubes Output

8. T-Test

Select "T-test" from the analysis menu shown in Fig. 2. The page shown in Fig. 13 is displayed.

Fig. 13 T-test
Fig. 13 T-test

Fig. 14 shows the output obtained when "TQ3_c: Work will become more efficient and easier" is selected for "the variable that you want to test" and "Q3_8: PC" is selected for "the variable that groups cases into two groups". This is the exact same type of output as obtained with "T-Test of Independent Samples" with the standard PC version of SPSS. The box at the top displays the variable name, case number (N), average value, standard deviation, and standard error of mean. The bottom box shows the items needed for a homogeneity of variance test (F value and significance probability), and the items for a T-Test for equality of means (T value, degree of freedom, significance probability, difference of mean, standard error of differences, and 95% confidence interval of the differences ).

Fig. 14 T-Test Output
Fig. 14 T-Test Output

9. One-way ANOVA

Select the dependent variables and factors (independent variables) from the list on the left. Three types of test can be performed - least significant difference, R-E-G-W Q, and Turkey’s b. The available options are descriptive statistics, homogeneity of variance test, and plot of mean. The default selections are least significant difference and Turkey’s b for the test and descriptive statistics and homogeneity of variance test for the options. Missing data is excluded from each analysis.

Fig. 15 One-way ANOVA
Fig. 15 One-way ANOVA

Fig. 16 shows the output when "Q3_8: Computer" is selected for the dependent variable, "Age10: 10YRS cohort" is selected for the factor variable, the verification and options savings left at the default settings, and the "ANOVA" at the bottom of the page clicked. A large amount of data is output, so only a portion of this data is shown in Fig. 16.

Fig. 16 One-way ANOVA (Partial)
Fig. 16 One-way ANOVA (Partial)

Following the "Descriptives", "Test of Homogeneity of Variance" and "ANOVA" shown in Fig. 16, the output data continues with a multiple comparison. The output results are the same as those obtained with the PC version of SPSS.

10. Bivariate correlations

Select "Bivariate Correlation "from the "Analysis" menu shown in Fig. 2 to open the screen shown in Fig. 17. Select the variables for which the correlations are to be checked from the list on the left side of the screen. Three analysis methods can be selected -- Pearson, Kendall, and Spearman. For the optional settings, "Two-tailed" or "One-tailed" can be selected for "Test of Significance", and "Exclude cases pairwise" or "Exclude cases listwise" can be selected for "Missing Values". For "Statistics", "Means and standard deviations" or "Cross-product deviations and covariances" can be selected.

Fig. 17: Bivariate correlations
Fig. 17: Bivariate correlations

For the variables, select "Q35: Personal income" and "EDU: Education YRS", leave the default settings for the test and analysis options, and then click the [Bivariate Correlations] button at the bottom of the screen to output the results. An example of this output is shown in Fig. 18. The output results are the same as the results obtained with the Windows version of SPSS. Values are output for the Pearson Correlation, significance (2-tailed), and number of cases (N).

Fig. 18: Output from Bivariate Correlation
Fig. 18: Output from Bivariate Correlation

11. Partial Correlation

Select "Partial Correlation" from the "Analysis" menu shown in Fig. 2 to open the screen shown in Fig. 19. Select the variables for which the correlations are to be checked and the variables to be controlled from the list on the left side of the screen. For the optional settings, "Exclude cases pairwise" or "Exclude cases listwise" can be selected for "Missing Values", and for "Statistics", "Means and standard deviations" or "Zero-order correlations" can be selected.

Fig. 19: Partial Correlation
Fig. 19: Partial Correlation

For the variables, select "Q35: Personal income" and "Q18: Perceived stratification", select "EDU: Education YRS" for the control variable, leave the default settings for the analysis options, and then click the [Partial Correlations] button at the bottom of the screen to output the results. An example of this output is shown in Fig. 20. The output results are the same as the results obtained with the Windows version of SPSS, although the display format is different. The control variables are output in the first column, and the Pearson Correlation coefficient, significance probability (2-tailed) and number of cases (N) are output in the third column.

Fig. 20: Output from Partial Correlation
Fig. 20: Output from Partial Correlation

12. Factor analysis

Select "Factor Analysis" from the "analysis" menu shown in Fig. 2 to open the screen shown in Fig. 21.

From the list on the left side of the screen, select the variables for which factor analysis is to be performed. The available factor extraction methods are "Principal Components", "Unweighted least squares method", "Generalized least squares method" , "Maximum likelihood method" , "Principal axis factoring method" , "Alpha factoring method", and "Image factoring method". The standard for factor extraction can be set to "Minimum eigenvalue" or "Num. of factors". For the optional settings, "None", "Varimax", or "Promax" can be selected for the rotation method. For factor extraction, "Correlation matrix" or "Variance-covariance matrix" can be selected, and for "Missing Values", "Exclude case listwise " or "Exclude case pairwise" can be selected. For the descriptive statistics, "Initial Solution", "Univariate descriptive statistic", "Coefficients", "Inverse matrix", "Significance levels", "Reproduced correlations", "Determinant", "Anti-image", or "KMO and Barlett's test of sphericity" can be selected.

Fig. 21: Factor Analysis
Fig. 21: Factor Analysis

For the variables, select "TQ11i: People must have respect for authority", "TQ11j: The only way to know what you should do is to rely on leaders and experts", "TQ11k: Sex criminals should be publicly whipped or worse", "TQ11l: People must obey their superiors", "TQ11m: A good leader has to be strict with his or her followers", and "TQ11n: People who question traditional and established ways are the cause of problems." Leave the default settings for the analysis options and then click the [Factor Analysis] button at the bottom of the screen to output the results. A portion of an example of output is shown in Fig. 22. The output results are the same as those obtained with the Windows version of SPSS, but here, the scree plot factor is also output (although it is not shown in this Fig.). The top block shows the output for the commonality of the initial values and the values after extraction, the center block shows the totals for the explained variances, and the bottom block is the output of the component matrix.

Fig. 22: One Portion of Factor Analysis Output
Fig. 22: One Portion of Factor Analysis Output

13. Linear Regression

Select "Linear Regression" from the “Analysis Menu” shown in Fig. 2. The page shown in Fig. 23 is displayed.

Fig. 23 Linear Regression
Fig. 23 Linear Regression

In addition to the selection of one dependent variable and independent variables, optional selections include the enter method ("Enter" or "Stepwise"), the type of statistic ("Model fit", "Descriptives" and/or "Collinearity diagnostics"), the use of a histogram for the standardized residual plot, and the missing data ("Exclude cases listwise" and " Exclude cases pairwise"). The default selections are" Enter", "Model fit", "Descriptives" and "Exclude cases listwise".

Fig. 18 shows a portion of the output when "Q8: How many days reading daily newspapers a week" is selected as the dependent variable, "Q1: Gender", "Q2: Age" and "EDU: Education YRS" are selected as the Independent variables, the options are left at their default selections, and the "Regression" button at the bottom of the page clicked. The actual output begins with "Descriptive Statistics", "Correlations" and "Variables Entered/Removed", and then continues with the items shown in Fig. 18 - "Model Summary", "ANOVA" and "Coefficients". These output results are also the same as the output obtained when using the PC version of SPSS.

Fig. 24 Linear Regression Output (Partial)
Fig. 24 Linear Regression Output (Partial)

14. Multinomial Logistic Regression

You can perform Multinomial Logistic Regression with categorical variables as dependent variables. The Multinomial Logistic Regression procedure with “Q6: Use of cellular phones” included in the Japan Survey on Information Society (JIS2004) dataset as a dependent variable is described here.

Procedure
  1. Click on Multinomial logistic regression on the Analysis Menu window. A new window as shown in Figure 23 will appear.
  2. Select the categorical variable Q6: Use of cellular phones as a dependent variable in the list of variables on the left side of Select Variables and click on the >> button. You can select a Reference category from among the three categories, which are First category, Last category, and Custom. Also you can select a Category order out of the two orders, which are Ascending and Descending orders. The defaults are First category and Ascending order, respectively. To change the dependent variable you use, click on the << button and you can change it back to the original.
  3. Of the independent variables you want to use in the analysis, input categorical variables in Factor(s) and quantitative variables in Covariate(s). As an example, select Q1: gender and Education 3 categories, click on the >> button, and proceed to the Factors column. Likewise, select age and Q39: Household income and proceed to the Covariate(s) column.
  4. The options you can select are: Case processing summary in Statistics; Pseudo R-square, Step summary, Model fitting information, Cell probabilities, Classification table, and Goodness-of-fit in Model information; and Estimates confidence interval (%), Likelihood ratio tests, Asymptotic correlations, and Asymptotic covariances in Parameters. The default settings are: Case processing summary, Pseudo R-square, Step summary, Model fitting information, Estimates confidence interval %: 95, and Likelihood ratio tests.
  5. In Convergence Criteria, you can provide settings for Iterations, Delta, and Singularity tolerance. The default settings are: Maximum iterations: 100, Maximum step-halving: 5, Log-likelihood convergence: 0, Parameter convergence: 0.000001, and Check separation of data points from iteration: 20 forward.
  6. After completing tasks i. to v., click on the Multinomial logistic regression button and Case Processing Summary, Model Fitting Information, Pseudo R-Square, Likelihood Ratio Tests, and Parameter Estimates will appear. Figure 24 shows a part of the results that appear on the screen. Refer to this information as necessary. In the analysis example given here, all settings are the defaults that remain unchanged.
Fig. 23 Select Variables, Statistics, and Convergence Criteria
Fig. 23 Select Variables, Statistics, and Convergence Criteria
Fig. 24 Multinomial Logistic Regression Output (a Part)
Fig. 24 Multinomial Logistic Regression Output (a Part)

15. Hierarchical Cluster Analysis

The Hierarchical Cluster Analysis procedure is described here using the data included in the Japan Survey on Information Society (JIS2004). In the SRDQ Hierarchical Cluster Analysis, a variable or a case can be selected for clustering. Note, however, that if the number of cases involves large amounts of data, and the Hierarchical Cluster Analysis is executed, it will take a considerable amount of time to process the data and the window may freeze. To avoid this, the SRDQ limits the objects of clustering in the Hierarchical Cluster Analysis to variables. If you perform cluster analysis for cases, you are advised to use the K-Means Cluster Analysis.

When you select the Hierarchical Cluster Analysis in the analysis menu shown in Fig. 2, the window in Fig. 25 will appear.

Fig. 25 Hierarchical Cluster Analysis
Fig. 25 Hierarchical Cluster Analysis

Select a variable you want to cluster in the Select Variables column in Fig. 26, click on the >> button, and proceed to the Variables column. For the analysis results, you can choose whether to indicate the Statistics or the Plots (Dendrograms).

Fig. 26 Select Variables
Fig. 26 Select Variables

In the Options in Fig. 27, you can choose whether to indicate Agglomeration schedule or Proximity matrix for the Statistics. (If you have chosen None in the Cluster membership below, you must put a check in Agglomeration schedule or Proximity matrix, or both without fail.)

For the Cluster membership, you can select None, Single solution or Range of solutions. The aim is to indicate to which cluster each object belongs, based on the results of cluster analysis. If you choose Single solution, you must enter the number of clusters and if you choose Range of solutions, you must enter the minimum number and the maximum number of clusters.

Fig. 27 Statistics
Fig. 27 Statistics

Then, as shown in Fig. 28, choose a Method for use in the cluster analysis as an option. You can select a Cluster method from among the cluster methods Between-groups linkage, Within-groups linkage, Nearest neighbor, Furthest neighbor, Centroid clustering, Median clustering, and Ward’s method.

For Measures, choose a method for measuring the distance between objects (similarity and nonsimilarity). The choice of this measuring method varies depending on whether data are measured in terms of Interval, Distribution, or Binary.

  • If data are measured in terms of Interval, you can select a measuring method from among the Squared Euclidean distance, cosine, Pearson correlation, Chebychey, Block, Minkowski, and Customized. If you choose Minkowski, you must choose the value of Power from among 1 to 4. If you choose Customized, you must choose the values of Power and Root from among 1 to 4, respectively.
  • If data are measured in terms of Distribution, you can select a measure from Chi-square measure and Phi-square measure.
  • If data are measured in terms of Binary, you can select a measure from among Euclidean distance, Squared Euclidean distance, Size difference, Pattern difference, Variance, Dispersion, Shape, Simple matching, Phi-4 point correlation, Lambda, Anderberg’s D, Dice, Hamann, Jaccard, Kulczynski 1, Kulczynski 2, Lance and Williams, Ochiai, Rogers and Tanimoto, Russel and Rao, Sokal and Sneath 1, Sokal and Sneath 2, Sokal and Sneath 3, Sokal and Sneath 4, Sokal and Sneath 5, Yules’s Y, and Yules’s Q. If you intend to select a binary method, you must enter the True and False values, respectively.

If you have selected Interval or Distribution for Measures, you can Standardize the measured values in the Transform values. You can select a standardization method from among None, Z scores, the range -1 to 1, the range 0 to 1, the maximum magnitude of 1, the mean of 1, and the standard deviation of 1. If you have selected methods other than None, you must choose between by each variable and by each case for conversion.

If you have chosen by each case, you can designate Absolute values, change sign, or rescale to 0-1 range.

Fig. 28 Method
Fig. 28 Method

An example of analysis is given below. In Select Variables, select Q1: Gender, age, Q4: Are you using a personal computer?, Q39: Your household’s revenue and your years of education, click on the >> button, and proceed to the Variables column. Leave the default options as they are, click on the Hierarchical Cluster Analysis button at the bottom of the page, and the results will be outputted. The outputs are shown in Fig. 29 and 30. First, a Case Processing Summary will be outputted. Then an Agglomeration Schedule will be outputted. Finally, a dendrogram will be outputted because Plots (Dendrograms) are chosen.

Fig. 29 Output of the Hierarchical Cluster Analysis (Case Processing Summary and Agglomeration Schedule)
Fig. 29 Output of the Hierarchical Cluster Analysis (Case Processing Summary and Agglomeration Schedule)
Fig. 30 Output of the Hierarchical Cluster Analysis (Dendrogram)
Fig. 30 Output of the Hierarchical Cluster Analysis (Dendrogram)

16. K-Means Cluster Analysis

The K-Means Cluster Analysis procedure is described here using the data included in the Japan Survey on Information Society (JIS2004). Unlike the Hierarchical Cluster Analysis, this analysis requires you to set the number of clusters in advance. It is suitable for handling large amounts of data involving numerous cases to analyze.

First, select K-Means Cluster Analysis in the Analysis Menu shown in Fig. 2, and a window will appear as shown in Fig. 31. Then, select a variable to be used in the analysis in Select Variables, click on the >> button, and proceed to the Variables column on the right. If you wish to use a variable as a Case Label, select it in the Select Variables column in the same way. Enter the number of clusters into which you want to classify objects in Number of Clusters (2 as default). For the classification Method, choose between Iterate and classify and Classify only.

On the right side of the window, define Options. For Iterate, enter the values of Maximum iterations and Convergence criterion (10 is the default for the maximum iterations and 0 is the default for the convergence criterion). You can set whether to Use running means or not.

In Save new variables, you can set whether to save the analysis results of Cluster membership and Distance from cluster center as variables or not.

In Statistics, you can set whether to indicate Initial cluster centers, ANOVA table, and Cluster information for each case as the analysis results or not.

In Missing values, choose between Exclude cases listwise and Exclude cases pairwise in the treatment of missing values.

Fig. 31 K-Means Cluster Analysis
Fig. 31 K-Means Cluster Analysis

An example is given below. For variables, select Q1: Gender and age, Q4: Are you using a personal computer?, Q39: Your household’s revenue and your years of education. For options, change only the default number of clusters to 4 and click on the K-Means Cluster Analysis button at the bottom of the page, and the results will be outputted. The outputs are shown in Fig. 32, 33, and 34. Initial Cluster Centers will be outputted first. Then Iteration History and Final Cluster Centers will be outputted. Finally, the Number of Cases in Each Clusters, the Number of Valid cases, and Missing values will be outputted.

Fig. 32 Initial Cluster Centers
Fig. 32 Initial Cluster Centers
Fig. 33 Iteration History
Fig. 33 Iteration History
Fig. 34 Final Cluster Centers and the Number of Cases in Each Cluster
Fig. 34 Final Cluster Centers and the Number of Cases in Each Cluster

17. Multi-way ANOVA

Using the data included in the Japan Survey on Information Society (JIS2001), the Multi-way ANOVA procedure is described below. Select Multi-way ANOVA in the Analysis Menu shown in Fig. 2, and the window as shown in Fig. 35 will appear.

In the Select Variables column, select one quantitative variable, click on the >> button, and proceed to the Dependent Variable column. Of the dependent variables you want to use in the analysis, input categorical variables in Fixed Factor(s) and quantitative variables in Covariate. Here, for a dependent variable, select Q18: Perceived stratification, and for fixed factors, select Q1: Gender and Education 3 categories.

In the Detailed condition, you can designate Model, Contrasts, Plots, Post Hoc, or Options. Put a check in each one of them, and a new window will appear for designating options at the bottom of the window shown in Fig. 35.

Fig. 35 Multi-way ANOVA 1
Fig. 35 Multi-way ANOVA 1
Fig. 36 Multi-way ANOVA 2
Fig. 36 Multi-way ANOVA 2

In Model, you can choose the type of Sum of squares from among Type I to Type IV. The default type of Sum of squares is Type III and Include intercept in model is checked.

In Contrasts, you can verify differences, if any, in Dependent Variables between the levels of variables designated in Fixed Factor(s). Select the factors you want to compare from among Factor(s), select a comparison method among Deviation, Simple, Difference, Helmert, Repeated, and Polynomial from the pull-down menu in the middle, click on the >> button, and proceed to the Set value column.

In Plot, the average value of a dependent variable for each factor that is estimated from the model can be graphically shown. In the Factor(s) column on the left side, Gender (q1) and Education 3 categories (edu_3) are indicated. Designate a variable to use for the horizontal axis in Horizontal Axis and variables to use in classifying lines in Separate Lines. To create a graph with the gender shown on the horizontal axis and the education 3 categories shown in a linear form, select “q1” in the Factor(s) column, click on the >> button, and proceed to Horizontal Axis. Likewise, select “edu_3,” proceed to Separate Lines, and then click on the Add button. In the Plot column on the right side, “q1*edu_3” is shown.

In Post Hoc Multiple Comparisons for Observed Means, provide settings for multiple comparisons (Fig. 37). Select a variable with which you want to perform multiple comparison, click on the >> button, and proceed to the Post Hoc Multiple Comparisons for Observed Means column, and check the multiple comparison method you use.

Fig. 37 Multi-way ANOVA 3
Fig. 37 Multi-way ANOVA 3

In Options, indicate means through Factor(s) and Factor Interactions and specify the indication of other statistics. In Display Means for, select factors whose means you want to verify and factor interactions in the Factor(s) and Factor Interactions column on the left side and proceed to the Display Means for column on the right side. As the message “If there are more than 5 fixed factors in the model, the interactions are not included” is given below the Factor(s) and Factor Interactions column, SRDQ Multi-way ANOVA sets a limitation that the interactions can be indicated only if there are less than four fixed factors. In Display below, it is possible to specify to display various options, such as Descriptive statistics and Homogeneity tests.

Click on the Multi-way ANOVA button at the bottom of the page, and the results will be outputted. There are many outputs, and only some of them are shown in Fig. 38.

Fig. 38 Partial Outputs of Multi-way ANOVA
Fig. 38 Partial Outputs of Multi-way ANOVA

18. Case Selection

Select "Select Cases" from the "Analysis" menu shown in Fig. 2 to open the screen shown in Fig. 39.

This section provides an explanation of the procedure used to select cases. First, select the variable name from the list at the bottom left of the screen and click the [?] button. The selected variable name is displayed in the input box at the top, under "Definition of IF Condition". Then, the same procedure can be used to select the variable value from the list on the bottom right, and the functions, operators, and numbers using the list and buttons on the right side of the screen to display these items in the input box at the top of the screen. In this way, a combination of variables, variable values, and operators can be displayed to create a definition of an IF condition in the input area at the top of the screen. When the IF condition definition has been created, click [OK] to apply the definition. If [Clear] is clicked followed by [OK], the created IF condition is cancelled.

The syntax used here for case selection is nearly the same as the syntax used for SPSS "Case Selection: Definition of IF Conditions". However, the case selection functions of this program also include a function by which, if the variable name is selected, the individual variable values are displayed, and the values of each can be selected. This function is not available with SPSS.

Fig. 39 Case Selection
Fig. 39 Case Selection

For this example, select the "Male" case. For the variable name, select "Q1: Gender", for the operator, select "=", and select "1.0: Male" for the variable value. This procedure creates the "q1=1.0" expression in the input box at the top of the screen. The "q1=1.0" expression can also be entered directly into the input box, without following this selection procedure. With "q1=1.0" displayed in the input box, click [OK]. With this operation, it is possible to apply the condition of "males only" to the data so that all subsequent analyses will be limited to male subjects only.

19. Recoding Values

Select "Recode" from the "Analysis" menu shown in Fig. 2 to open the screen shown in Fig. 40.

This section shows the procedure for recoding values. In the "Select an Input Variable" list on the left, select the variable for which the value is to be recoded and then click the [>>] button to add the variable to the "Input Variables >> Output Variable" list on the right. Next, enter the "Name (Output Variable)" and "Label (Output Variable)" and click the [Define] button to create the new variable name and label. Lastly, use the following procedure to recode the values. At the bottom of the screen, enter the old value, new value and new value label and then click the [Add] button. The corresponding values are shown under "Old >> New" at the bottom right of the screen. For the old value, in addition to the value, a missing value and value range can also be specified. With the recoding conditions for the corresponding value shown under "Old >> New", click the [OK] button. When this is done, a new variable with the recoded value will be added to the list, and subsequent analyses can be conducted using the new variable. When the [IF] button located at the center of the right side of the screen is clicked, a screen for creating IF statement conditions, like the one shown in Fig. 39) is displayed. Following the procedure already described for defining IF conditions, the screen can be used to add IF conditions, and value recoding can be performed after the new conditions have been added. The syntax used for recoding variable values is the same as that used for SPSS "Recoding Values to Other Variables".

Fig. 40 Recoding Variable Values
Fig. 40 Recoding Variable Values

Here, the input variable "AGE10: 10YRS cohort" is converted to three categories, "under 50", "from 50 under 70", and "70 and up", and the "age 3 category" variable is created. First, select "Q2* Age: By 10 Years" under "Select an Input Variable", move the variable to "Input Variables >> Output Variable", enter "age3c" for the output variable name and "age 3 category" for the output variable label, and click the “Define” button. Then, using the values applied thus far, select "Range:", enter "1" and "3" for the range values, enter "1" for the new value and "under 50" for the new value label, and then click the [Add] button. Likewise, using the values applied thus far, enter "4" and "5" for the range values, enter "2" for the new value and "from 50 under 70" for the new value label, and click the [Add] button, and then, using the values applied thus far, enter "6" and "7" for the range values, enter "3" for the new value and "70 and up" for the new value label, and click the [Add] button. Lastly, click the [OK] button to create the new "age 3 category" variable for the data.

The command for recoding the created value can be displayed in command format, and value recoding can also be implemented by editing this command. Click the [command-line] button in the recoding screen shown in Fig. 40 to open the screen shown in Fig. 41.

Fig. 41 Directly Editing Recoding Commands
Fig. 41 Directly Editing Recoding Commands

The commands that are currently applied are shown in the box at the top of the screen. If the value for the "age 3" category shown above is being recoded, "age3c:age10:1 thru 3=1,4 thru 5=2,6 thru 7=3:age 3 category:1=under 50,2=from 50 under 70,3=70 and up:" is displayed. Please be aware that the syntax of this command differs from the command syntax used with SPSS. The meanings of each command item are as follows: (New variable name) : (Old variable name) : (Old variable value) = (New variable value) : New variable label) : (New variable value) = (New variable label)

To edit the command and implement a new recoding, copy the command shown in the box at the top, paste it into the input box at the bottom of the screen, and edit the command.

For the edited command, we will change the age breakpoint from 50 to 40. After copying the command from the box at the top, change "1 thru 3=1,4 thru 5=2" to "1 thru 2=1,3 thru 5=2", and change "1=under 50,2=from 50 under 70,3=70 and up:" to "1=under 40,2=from 40 under 70,3=70 and up:". Verify that "age3c:age10:1 thru 2=1,3 thru 5=2,6 thru 7=3:age 3 category:1=under 40,2=from 40 under 70,3=70 and up:" is shown for the command in the bottom box and then click the OK button to create an age 3 category with an age breakpoint of 40.

The advantage to using the command editing procedure is that a newly created variable command can be saved. With recoding and variable calculation, even if new variables are created, those variables cannot be saved for use when the next analysis is performed. However, when newly created variable commands are saved as text data, when the next analysis is performed, a variable that was created during a previous analysis can be used again simply by pasting the command into the editing box and clicking the OK button.

20. Computing Variables

Select “Compute Variables” from the “Analysis” menu shown in Fig. 2 to open the screen shown in Fig. 42.

This section describes the procedure for computing variables. First, enter the name and level for the target variable and click the Add button; the variable label is then displayed in the box at the top left of the screen. Next, create the numeric expression in the input box at the center of the screen. When a variable name is selected, that name is displayed in the numeric expression. In addition to variable names, functions, operators, and numbers can also be selected, at which point they are displayed in the expression. After using this procedure to complete the numeric expression, click the “Define” button; the target variable and calculation expression are then displayed in the box at the top of the screen. The calculation expression can also be entered directly. Lastly, click the OK button to create the target variable. When the IF button located under the variable level input box is clicked, the IF options screen shown in Fig. 23 opens. IF conditions can be added using the same procedure described for defining IF conditions, and variables can be computed with the IF conditions applied to the computation. The variable computation used here is the same as SPSS variable computation, although the display format varies slightly.

Fig. 42 Variable Computation
Fig. 42 Variable Computation

Here, we will create a variable for the total number of information instruments. First, enter “infonum” for the variable name and “number of information instruments” for the variable label and click the Add button. Then, under “Variables”, select “Q3_8: PC”, select “+” for the operator, and then continue, selecting “Q3_9: Printer” “+” “Q3_10: Scanner” “+” and “Q3_11: Digital camera” to create the expression “q3_8+q3_9+q3_10+q3_11”. Lastly, click the “Define” button to display the created expression under “Target variable:”, and then click the [OK] button. Through this procedure, the new “number of information instruments” variable is created and can be used in analysis.

In the next example, we create a category for full-time housewives, which will be applied only to female subjects. Enter “syufu” for the variable name and “Full-time housewife dummy” for the variable label, and then click the “Add” button. Next, click the “IF” button located under the “Add” button and add the condition to limit the subjects to females only. Following the procedure described in section “13. Case selection”, select “Q1: Gender” from the list of variables on the left, select “=” for the operator, and then select “2.0: Female” from the list of values on the right. Verify that “q1=2.0” is shown for the IF condition and then click the [OK] button. With this procedure, the “q1=2.0” condition is added to the IF item of the Compute screen. Lastly, add two additional conditions needed to create the final “full-time housewife” variable- the subject must be married, and, must be unemployed. Under “Variables”, select “Q33: Marital status”, select “=” for the operator, and then select “2.0: Married (has a spouse)” under “Values” to create the “q33=2.0” condition. Also, select “Occupation 8 categories” under “Variables”, select “=” for the operator, and then select “9.0: Unemployed” under “Values” to create the “occ8=9.0” condition. Both of these conditions must apply in order for the subject to be a full-time housewife, so each condition must be placed in parentheses and the two conditions connected by the ampersand (“&”) operator to create the “(q33=2.0)&(occ8=9.0)” expression. Click the [Define] button to display the target variable, and then click the [OK] button. By creating this variable, the “Full-time housewife dummy” variable is created under the set of conditions applicable to female subjects, and this variable can be used in analysis.

The command for computing the created valuables can be displayed in command format, and computing valuables can be implemented by editing this command. In the “Variable Computation" screen shown in Fig. 42, click the [command-line] button to open the screen shown in Fig. 43.

Fig. 43 Directly Editing Compute Commands
Fig. 43 Directly Editing Compute Commands

The box at the top of the screen shows the commands that are currently applied. If a computation is made using the “number of information instruments” variable described earlier, “infonum:q3_8+q3_9+q3_10+q3_11:number of information instrument:” is shown in this box. The command format indicates “(new variable name): (numeric expression): (new variable label)”. To copy this command and compute the variable, copy the command shown in the top box, paste the command into the box at the bottom of the screen and edit the command.

Say, for example, that the corresponding information instrument should be “Q3_1: Cellular phone” rather than “Q3_11: Digital camera”. In the command copied from the top box, replace “q3_11” with “q3_1” so that “q3_8+q3_9+q3_10+q3_11” becomes “q3_8+q3_9+q3_10+q3_1”. Verify that “infonum:q3_8+q3_9+q3_10+q3_1:number of information instrument:” is shown in the bottom box and then click the OK button. This creates a variable for which the number of information instruments includes cellular phones rather than digital cameras.