To:  sbaker@odu.edu

From:  Thomas Meyer
tmeyer@ph.vccs.edu
276 656-0283
Patrick Henry Community College

Subject:  Statistics - w/  Dr.Spencer Baker - Midterm Exam - Due date, March 1,2004.

Date: February 28, 2004

                                            

Procedure followed:


Click the following links:
 
Step # 1.       
 Looking at the data
 

Tables

1 -    Variable View from SPSS
 2A -  Kurtosis and Skewness of Japanese Variables
2B -  Kurtosis and Skewness of American Variables
3A -  Statistics and Histograms about USA Respondents
3B -  Statistics and Histograms about Japanese Respondents
 4 -     Boxplot Comparisons Between Japanese Respondents and  USA Respondents
 5A -  Correlations Among Variables Involving Japanese Respondents
5B - Correlations Among Variables Involving American Respondents

 Based on this information, report whether or not you have concerns about conducting the analysis
and why.
 
Step # 2.       
Selecting & Conducting
appropriate analysis
 
Step # 3.       
Interpreting a Results Section

Tables and Figures
 Statistical Presentation
 Inferences
 

References

 

 


Step #1 - Looking at the data

Notes to accompany Table 1
1.  Each of 10 named variables are numeric and represented by strings of letters from three to seven characters in length. 
2.  Variables are each given more explanatory labels.
3.  In the gender variable, females are coded 1; males are coded 2.
4.  In the citizen category, participants from Japan are coded 1; participants from the USA are coded 2.
5.  Age. gender. and citizen are measured as ordinal variables; the remaining seven variables are measured as scale variables.
6.  None of the data are reported as missing.  None of the data have decimal values reported.
7.  There are 199 "cases" or persons reporting data on each of these ten variables.
8.  Ages of respondents range from 17 to 24 years old.
9.  94 "cases" or respondents are reported as having Japanese citizenship.
10.  105 "cases" or respondents are reported as having United States of America citizenship.

Table 1 - Variable View from SPSS
 
Variable
Name		 Type Width Decimals Label Values Missing Columns Align Measure
age Numeric 3 0 Age of the Participant None none 8 Right Ordinal
gender Numeric 6 0 Gender of the Participant 1=Female;
2=Male		 none 8 Right Ordinal
citizen Numeric 7 0 Country of Citizenship 1=Japan; 
2= USA		 none 8 Right Ordinal
collect Numeric 7 0 Collectivism Scale Score none none 8 Right Scale
ioam Numeric 4 0 Individually-Oriented Achievement Motivation none none 8 Right Scale
soam Numeric 4 0 Socially-Oriented Achievement Motivation none none 8 Right Scale
indse Numeric 5 0 Individualistic Self-Esteem none none 8 Right Scale
cses Numeric 5 0 Collectivistic Self-Esteem none none 8 Right Scale
indsc Numeric 5 0 Independent Self-Construal none none 8 Right Scale
intsc Numeric 5 0 Interdependent Self-Construal none none 8 Right Scale

 

                                   

Based on this information, report whether or not you have concerns about conducting the analysis and why.

Answer: 
Ten items noted below bring strong concerns to my mind as a researcher:

1.  Who are the respondents?
Respondents are reported as 17 to 24 years old.  What we don't know is whether these respondents are the current "college age crowd" of Japan and America, or whether these persons were pre-World War II 17 to 24 year-olds.  The values held prior to the Emperor of Japan asking his people to cease hostilities toward the United States are likely very different from values held in society in Japan today.
2.  Can the respondents be taken as a microcosm for their respective societies? 
Respondents 17 to 24 years old do not necessarily represent the rest of societal values in either Japanese or American values.
3.  How were the respondents chosen?
The study makes no reference to random selection of participants, further invalidating the ability to generalize results to the respective cultures.  The absence of random selection of participants presents severe limitations upon the manipulation of the data in ways that would require randomization.
4.  What about the possibility of a respondent possessing dual citizenship?
The results of the study could be impaired by poor boundary maintenance.  A respondent can potentially, with an American and a Japanese parent, have dual citizenship.  These " dual citizens" could show up as respondents in both countries, adding a certain "noise" by reducing the discriminatory power of the country of citizenship variable.
5.  What restrictions will a convenience sample impose upon statistical procedures? 
The study "appears" to consist of a "convenience sample" conducted on campuses in either the USA, Japan, or both countries.  No amount of manipulation of data can overcome the limitations imposed by a non-random sample.
6.  Is there a form of response bias built into the study?
The fact that there are 199 respondents (and not 200) raises the question as to whether it is likely that the original formulators of the study rejected one case arbitrarily.  This would be a form of response bias.
7.  Can barriers imposed by language differences be overcome? 
Instruments administered to Japanese and American citizens may have language barriers to overcome.  Nuances in one language may not be present, or at least be different in a second language.  Some expressions in one language may not be constructs readily apparent in a second language.   It is unknown whether respondents were using instruments with equivalent meanings.  It is unknown if the measurement devices were given in the native language of the respondent.  It is not even reported what the measurement devices consisted of. 
8.  What has been done, given language differences in Japan and America,  to address the problem of inter-rater reliability?
It is likely that whether the measurement instruments be either qualitatively or quantitatively different, the condition in which two of more researchers using the same measures with the same study participants would get the same result needs to be addressed.
9.  Can differences in frames of reference unique to two different cultures be bridged?
Also unknown is the extent to which the instruments and their language convey the superiority of one culture over another, or whether the instruments of measurement manifest an equivalency of the two cultures.
10.  Was the experimental design so flawed at its outset as to render statistical results invalid?
No amount of manipulation of the data can overcome deficiencies in experimental design.  Probabilities and confidence intervals cannot be established without further facts as to the randomization of selection.  This curtails our analysis to bivariate analysis - essentially to correlations and to comparisons of boxplots regarding responses made by persons whose citizenship differs.

 

 


 


Step 2 - Selecting & Conducting
appropriate analysis

From an examination of the data, we immediately see that the first three variables, Age, Gender, and Country of Citizenship are ordinal data.  The mode (highest frequency) is the only measure of central tendency to be of value with these three ordinal variables.

Seven variables possess scale, or interval measurement properties.  They are:
 
Collectivism Scale Score
Individually-Oriented Achievement Motivation
Socially-Oriented Achievement Motivation
Individualistic Self-Esteem
Collectivistic Self-Esteem
Independent Self-Construal
Interdependent Self-Construal

What's going on here?
It makes sense to separate the respondents by Country of Citizenship, in order to try to discover -
Are the response patterns in the seven scaled variables similar in American and in Japanese respondents?

How can we discover and measure what's happening?
If the first 94 Japanese respondents are to be compared to the remaining 105 American respondents on each variable,
we can examine and compare responses between American and Japanese respondents on each variable
by ordinarily employing the following numbered methods:

1.  Means, modes, medians - measures of central tendency
It is readily possible to calculate the average, the most frequently occurring, and the value above and below which 50% of the reported values in the frequency distributions lie.
2.  Variances and standard deviations - measures of variability
It is readily possible to calculate the average of the sum of the squared differences of the observed values from their mean for each scaled variable, known as the variance.  It is readily possible to also find the square root of the variance, known as the standard deviation for each variable.
3.  Boxtplots
Using SPSS create a series of boxplots of responses by Americans to each variable, and beside those boxplots create a second series of boxplots of responses by Japanese to each variable.  In this way, medians, interquartile ranges, and minumum and maximum response values on each scalable variable can be compared from one culture to the other.
4.  Bivariate correlation tables
Do a bivariate correlation using SPSS creating a table of ten variables by ten variables. 
     Positive correlations will tend to show an underlying upward sloping, direct, or positive relationship between variables.  The strength of the linear relation varies from a low of "zero" to a perfectly linear correlation value of 1.00.  (For example, each variable is perfectly correlated with itself, producing a diagonal line possessing correlation values of 1.00.)
     Negative correlations will tend to show an underlying downward sloping, inverse, or negative relationship between variables.  The strength of the linear relation varies from zero to a perfectly negative linear correlation value of -1.00.
5.  Frequency tables and histograms
We can create a frequency response on each scaled variable by American and by Japanese respondents.  Beside each frequency response we can create the histogram representing those response frequencies.  The relative frequencies  would have the same shape, but a different scale on the vertical axis.  And the relative frequencies could be used to imply that areas beneath the curve could represent the probablilities or chance of such events occurring if the response patterns seemed to contain the properties of a normal distribution.  (The normal distribution contains an equivalent mean, mode, and median, and while this might not be the observed results, we could gain insight into the general shape of the response patterns by seeing if because they are human characteristics, they seem to follow a normal distribution.)
6.  Skewness and kurtosis
The frequency tables and the histograms readily lend themselves to measures of skewness and kurtosis.  Skewness and kurtosis allow us to measure the extent to which each frequency response pattern conforms to or differs from the desired normal distribution by way of clustering responses in either tail, or by concentrating such responses near the center of the respective frequency table.
7.  t-tests for statistically significant differences between sample means
There may be grave difficulty in attempting to make t-tests between the American mean and the Japanese mean on each interval variable.  (Matters may not be ordinary here.)  We have no firm basis for believing that the samples of respondents were randomly chosen.  Were we to make such t-tests, we could not generalize to say that the underlying populations of Japanese and American peoples were being appropriately compared as to their different responses along the seven scaled variables.  We could only infer that the t-tests would purport to convey information about the populations the samples actually represent - which are likely to be, the students attending a particular school in America, and the students attending a particular school in Japan.  For the underlined reason mentioned, the t-tests will not be conducted.
8.  Power, effect size, and statistical significance
Due to the lack of random sampling, we lack the necessary prerequisite data to meaningfully assess power, effect size, and statistical significance.

 


 

 


Step 3 - Results Section

 

  Use these Links to:

Tables and Figures

Statistical Presentation

Inferences

 

 Results Section:   Tables and Figures
 

 

Tables and Figures

 

 

 

 

 
Variable View from SPSS    Use these links
   to see or to compare:
Table 1 -
Variable View from SPSS

 
Skewness and kurtosis

 

 

Table 2A - 
Kurtosis and Skewness of Japanese Variables		 Table 2B - 
Kurtosis and Skewness of American Variables

 

Means, modes, medians
- measures of central tendency

Variances and standard deviations - measures of variability

Frequency tables

Histograms
 

  
Table 3A -
Statistics and Histograms about USA Respondents		 Table 3B -
Statistics and Histograms about Japanese Respondents
Boxtplots
Table 4 -
Boxplot Comparisons Between
Japanese Respondents and  USA Respondents
 
Bivariate correlation tables
 
Table 5A -
Correlations
Among Variables Involving Japanese Respondents		 Table 5B -
Correlations
Among Variables Involving American Respondents
 
 

 

 

 
Notes to accompany Table 2A and Table 2B


1.  (The data were taken from homework problem 21 in Chapter 4.)
2.  Tables 2A and 2B show Kurtosis and skewness on the seven scaled variables.
3.  None of the values shown in Table 2A and Table 2B are extreme.  Extreme values (+ or - 2.0 deviations) would cause the
researcher to feel that the resulting distributions were so influenced by outlying values as to be rendered not subject to the
normal distribution.
4.  The impression of normality is further confirmed by histograms showing resemblance to normal distributions in each of the seven scaled variables.
5.  The meaning of kurtosis and skewness are as follows:
                            Interpretation of Kurtosis
 - A kurtosis of K = 0 is that assigned to a normal bell shaped distribution.
   - A negative number for K indicates  a flat or platykurtic distribution;
   - A positive number for K indicates a leptokurtic or highly peaked distribution.		                        
                     Interpretation of Skewness
    Pearson's index of skewness was computed on the seven scaled variables:

     -  will be positive for positively skewed distributions
        (having a long thin tail to the right)
     - will be negative for negatively skewed distributions
        (having a long thin tail to the left)
     -  Larger numbers indicate more severe skewness.


Table 2A -  Kurtosis and Skewness of Japanese Variables
 
JAPANESE RESPONDENTS 
Data on the 94 cases from Japan follow:

Variable Name        Kurtosis                                                           Skewness
                                           

1. collect :                Kurtosis =  .091   ;                                          Skewness =  -.225 
                                 Std. error = .493   ;                                            Std. error =  .249 
                                 leptokurtic (peaked)        negativelyskewed distributions (having a long thin tail to the left)

 2. ioam :                   Kurtosis = -.117    ;                                           Skewness =  .027
                                  Std. error =.493    ;                                             Std. error = .249 
                                   playtykurtic (flat)             positively skewed distributions (having a long thin tail to the right)
 
3. soam :                  Kurtosis =  -.427   ;                                            Skewness =  -.119
                                   Std. error = .493   ;                                             Std. error = .249 
                                   playtykurtic (flat)             negatively skewed distributions (having a long thin tail to the left)

4. indse :                  Kurtosis =  -.107   ;                                             Skewness =  .275
                                   Std. error = .493   ;                                              Std. error =  .249
                                   playtykurtic (flat)             positively skewed distributions (having a long thin tail to the right)

5. cses :                    Kurtosis =  1.119   ;                                            Skewness =  -.775
                                   Std. error = .493   ;                                              Std. error =  .249 
                                    leptokurtic (peaked)     negatively skewed distributions (having a long thin tail to the left)

6. indsc :                   Kurtosis =  -.105   ;                                             Skewness =  -.033
                                    Std. error = .493   ;                                              Std. error =   .249
                                    playtykurtic (flat)            negatively skewed distributions (having a long thin tail to the left) 

7. intsc :                     Kurtosis =  .206   ;                                              Skewness =  -.492
                                     Std. error = .493   ;                                             Std. error =  .249
                                     leptokurtic (peaked)    negatively skewed distributions (having a long thin tail to the left)

 

 


Table 2B -  Kurtosis and Skewness of United States Variables
 

UNITED STATES RESPONDENTS
Data on the 105 cases from United States of America follow:

Variable Name      Kurtosis                                                         Skewness

1. collect :               Kurtosis =  -.006   ;                                          Skewness =  -.287 
                                 Std. error = .467   ;                                            Std. error =  .236 
                                 playtykurtic (flat)              negatively skewed distributions (having a long thin tail to the left)

 2. ioam :                   Kurtosis =  .074    ;                                           Skewness =  -.508
                                  Std. error =.467    ;                                             Std. error = .236 
                                   leptokurtic (peaked)      negatively skewed distributions (having a long thin tail to the left)
 
3. soam :                  Kurtosis =   .174   ;                                            Skewness =  -.203
                                   Std. error = .467   ;                                             Std. error = .236 
                                   leptokurtic (peaked)      negatively skewed distributions (having a long thin tail to the left)

4. indse :                  Kurtosis =   .204   ;                                             Skewness =  -.830
                                   Std. error = .467   ;                                              Std. error =  .236
                                   leptokurtic (peaked)      negatively skewed distributions (having a long thin tail to the left)

5. cses :                    Kurtosis =  -1.72   ;                                            Skewness =  .633
                                   Std. error = .467   ;                                              Std. error =  .236 
                                    playtykurtic (flat)            positively skewed distributions (having a long thin tail to the right)

6. indsc :                   Kurtosis =  -.233   ;                                             Skewness =  .240
                                    Std. error = .467   ;                                              Std. error =   .236
                                    playtykurtic (flat)            positively skewed distributions (having a long thin tail to the right) 

7. intsc :                     Kurtosis =  .296   ;                                              Skewness =  -.233
                                     Std. error = .467   ;                                             Std. error =  .236
                                     leptokurtic (peaked)    negatively skewed distributions (having a long thin tail to the left)

 


 

 


Table 3A - Statistics and Histograms about USA Respondents
 
Frequencies 105 USA Respondents

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 


Table 3B - Statistics and Histograms about Japanese Respondents
 

Frequencies 94 Japanese Respondents

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 


Table 4 - Boxplot Comparisons Between Japanese Respondents and  USA Respondents
 


Japan                                                                                                                United States

Collectivism Scale Scores
compared below

 

 


Japan                                                                                                                United States

Individually-Oriented Achievement Motivation
 compared below

 

 

 

Japan                                                                                                                United States

Socially-Oriented Achievement Motivation
 compared below

 

 

 

Japan                                                                                                                United States

Individualistic Self-Esteem
 compared below

 

 

 

Japan                                                                                                                United States

Collectivistic Self-Esteem
 compared below

 

 

Japan                                                                                                                United States

Independent Self-Construal
 compared below

 

 

 

Japan                                                                                                                United States

Interdependent Self-Construal
 compared below

 

 


 

 


Table 5A - Correlations Among Variables Involving Japanese Respondents
 


 

Table 5B - Correlations Among Variables Involving American Respondents
 

 

 


 Results Section:   Statistical Presentation

 

Effect size, Power, and Degrees of Freedom
Effect size, Power, and Degrees of Freedom were not used.  The likelihood of non-random sampling negated the ordinary use of the t-test to determine the difference between means of the samples.

Statistical Significance
The following two tables do discover statistical significance between variables among the Japanese respondents, and then among the American respondents.

Analysis Used
A bivariate correlation was accomplished between variables for the Japanese respondents and then for the American respondents.


 
For respondents from Japan, using Pearson correlation coefficients, we can discuss:
    - the direction (positive or negative),
    - the extent (0=no correlation; 1=perfect correlation)
    - and the significance of each of the seven scaled variables with each other.
     * = Correlation is significant at the 0.05 level in a two-tail test.
    ** = Correlation is significant at the 0.01 level in a two-tail test.

For 94 Japanese respondents, the following statistically significant correlations occurred between variables:
Collectivistic Self-Esteem and Gender of the Participant: 
-.341** Pearson Correlation
.001 Sig. (2-tailed)
94 N

 

Collectivism Scale Score and Socially-Oriented Achievement Motivation
.250* Pearson Correlation  
.015 Sig. (2-tailed)  
94 N  

 

Collectivism Scale Score and Collectivistic Self-Esteem
.376** Pearson Correlation  
.000 Sig. (2-tailed)  
94 N  

 

Collectivism Scale Score and Independent Self-Construal
-.251* Pearson Correlation  
.015 Sig. (2-tailed)  
94 N  

 

Collectivism Scale Score and Interdependent Self-Construal
.506** Pearson Correlation  
.000 Sig. (2-tailed)  
94 N  

 

Individually-Oriented Achievement Motivation and Individualistc Self-Esteem
.243* Pearson Correlation  
.018 Sig. (2-tailed)  
94 N  

 

Individually-Oriented Achievement Motivation and Independent Self-Construal
.341** Pearson Correlation  
.001 Sig. (2-tailed)  
94 N  

 

Socially-Oriented Achievement Motivation and Collectivistic Self-Esteem
.226* Pearson Correlation  
.029 Sig. (2-tailed)  
94 N  

 

Socially-Oriented Achievement Motivation and Interdependent Self-Construal
.382** Pearson Correlation  
.000 Sig. (2-tailed)  
94 N  



For 105 American respondents, the following statistically significant correlations occurred between variables:

For respondents from America, using Pearson correlation coefficients, we can indicate:
    - the direction (positive or negative),
    - the extent (0=no correlation; 1=perfect correlation)
    - and the significance of each of the seven scaled variables with each other.
     * = Correlation is significant at the 0.05 level in a two-tail test.
    ** = Correlation is significant at the 0.01 level in a two-tail test.

 

Gender of the Participant and Individually-Oriented Achievement Motivation
-.292** Pearson Correlation    
.003 Sig. (2-tailed)    
105 N    

 

Gender of the Participant and Collectivistic Self-Esteem
-.240* Pearson Correlation  
.014 Sig. (2-tailed)  
105 N  

 

Age of the Participant and Independent Self-Construal
-.275** Pearson Correlation  
.005 Sig. (2-tailed)  
105 N  

 

Collectivism Scale Score and Independent Self-Construal
-.214* Pearson Correlation  
.028 Sig. (2-tailed)  
105 N  

 

Collectivism Scale Score and Interdependent Self-Construal
.482** Pearson Correlation  
.000 Sig. (2-tailed)  
105 N  

 

Individually-Oriented Achievement Motivation and Individualistic Self-Esteem
.362** Pearson Correlation  
.000 Sig. (2-tailed)  
105 N  

 

Individually-Oriented Achievement Motivation and Collectivistic Self-Esteem
.403** Pearson Correlation  
.000 Sig. (2-tailed)  
105 N  

 

Individually-Oriented Achievement Motivation and Independent Self-Construal
.361** Pearson Correlation  
.000 Sig. (2-tailed)  
105 N  

 

Socially-Oriented Achievement Motivation and Interdependent Self-Construal
.293** Pearson Correlation  
.002 Sig. (2-tailed)  
105 N

 

 

 


 Results Section:   Inferences
 
 

 

What does the analysis inform us of?
This analysis points to the critical nature of choosing samples in a random manner, and to anticipating design flaws before setting out to gather data.

 

Does the analysis support or not support the hypotheses?
The primary hypothesis was: "The collectivism scale score will have a statistically stronger relationship to individuals from Japan as compared to individuals from the U.S.A."

A t-test was not performed due to the likelihood of non-random sampling.  Therefore the boxplots provide a different means of comparing collectivism scale scores between the Japanese and American respondents. 

By observing the box plots of Japanese and American respondents side by side against collectivism scale score the following four numbered observations seem appropriate:

1.  Neither the ranges, medians, nor interquartile ranges differ significantly between respondents of either country.
Therefore the hypothesis is not supported by the currently reported data and respondents.
 


 

The secondary hypotheses were: "Measures of socially-oriented achievement motivation, collectivistic self-esteem, and interdependent self-construal
would be higher for citizens of Japan
than measures of individually-oriented achievement motivation, individualistic self-esteem, and independent self-construal."

A t-test was not performed due to the likelihood of non-random sampling.  Once again, boxplots provide a means of comparing the Japanese respondents on each of these variables to themselves.
 


2.  Contrary to expectation, the boxplot of Japanese individually-oriented achievement motivation has higher max, min, mean, and interquartile range than does the boxplot of  Japanese socially-oriented achievement motivation.

3.  In accordance with the expectation of the author of the study, the boxplot of Japanese collectivistic self-esteem did possess higher max, min, mean, and interquartile range than did the boxplot of Japanese individualistic self-esteem.

4.  Contrary to expectation, the boxplot of Japanese independent self-construal has higher max, min, mean, and interquartile range than does the boxplot of Japanese interdependent self-construal.

Secondary Findings

The following secondary findings result from comparing the boxplots of Japanese with American respondents:

5.  Japan respondents report slightly higher scores on the Collectivism Scale Score than do American respondents.

6.  Japan respondents report slightly lower scores on the Individually-Oriented Achievement Motivation measure than do American respondents.

7.  Japan respondents report slightly lower scores on the Socially-Oriented Achievement Motivation measure than do American respondents.  (This is contrary to common expectation.)

8.  Japan respondents report slightly lower scores on the Individualistic Self-Esteem  measure than do American respondents.

9.  Japan respondents report slightly lower scores on the Collectivistic Self-Esteem measure than do American respondents.  (This is contrary to common expectation.)

10.  Japan respondents report a slightly mean score on the Independent Self-Contrual measure than do American respondents.  (This contrary to common expectation.)  However the range of the American response extends to higher score values than does the Japanese range, and the majority of the American interquartile range report score values higher than their Japanese counterparts.

11.  Japan respondents report slightly scores on the Interdependent Self-Construal measure approximately equal to those scores reported by American respondents.  However, greater numbers of the interquartile range of Japanese respondents lie at lower scores than do their American counterparts.

12.  Reference source #1 mentioned beneath my signature contains the opinion of a San Francisco State psychologist that the current youth of Japan are quite different from the cultural expectations held by their parents, and by persons living in Japan prior to World War II.  The source provides a rationale as to why the findings in the current study may not be those initially expected.

 

Finis

 


filename: ODUstatMidTermExamTomMeyer.doc

Tom Meyer

Thomas Meyer

 

References:
1
This link summarizes psychology professor David Matsumoto (of San Francisco State University) who asserts that the current Japanese culture - collectivist, conscious of the needs of others, dedicated to their jobs, etc., is simply not supported by the most recent data.

  http://www.findarticles.com/cf_dls/m0NTN/2003_Jan/104732898/p1/article.jhtml

2
Website discussing O. J. Simpson trial in terms of "power" in the context used by statisticians.

  http://trochim.human.cornell.edu/OJtrial/oj4.htm
3
Notes about effect size measures.

  http://www.chsbs.cmich.edu/k_han/psy511/d.htm

 
4

Algorithims used from the homework:

1.  How to make a table of summary statistics about Japanese and about American respondents.

2.  How to create histogram charts

3.  How to make a vertical Boxplot:

How to do it, step by step:

Load the CD containing data;
Load SPSS;
Double-click SPSS;
File - Open - A-drive - Self.sav; Press Enter

Select Graphs menu
Select Boxplot... ; Define
Scroll to and select each variable, such as Collectivism Scale Score; click the arrow labeled Variable; (this puts Collectivism Scale Score on the vertical axis)
Scroll to and select Country of Citizenship; click the arrow labeled Category Axis;
Click OK

Right-click the picture; Copy;
Select a location on Frontpage; Paste
Save.
Refresh the web.

Click the left green arrow (to bring the menus back to active status)

Select Graphs menu
Select Boxplot... ; Define;  Click Reset;
Scroll to and select Collectivism Scale Score; click the arrow labeled Variable;
Scroll to and select Country of Citizenship; click the arrow labeled Category Axis;
Click OK

Right-click the picture; Copy;
Select a location on Frontpage; Paste
Save.
Refresh the web.