Home Page Overview
New Features Downloads & Prices
P-STAT documentation Technical Support
Contact P-STAT Unistat for Windows

101 Reasons for using the P-STAT SURVEY command

    The Variables

  1. Character variables. Are ordered alphabetically. Can be used as row (stub) or column (banner) variables.
  2. Numeric categorical variables. The values dictate the order. Can be used as either row or column variables.
  3. Numeric quantity variables The values are totaled and averaged and appear as one or two columns in the table.
  4. BY variables, if provided, cause each table to be produced for every combination of the values of the BY variables.
  5. A group of variables can be treated as a single multiple response variable.
  6. Multiple response variables can be either dummy (0/1) or multi-value variables.
  7. Multiple response variables can define either the columns (banners) or row (stubs) of a table.

    The Labels

  8. Extended labels for column variables up to 78 characters
  9. Extended labels for the row variables: an indefinite number of 78 character labels to contain the entire question or item
  10. Column (banner) labels up to 64 characters
  11. Optional break points for column labels may be supplied.
  12. Row (stub) labels up to 80 characters
  13. Labels can replace programmed text such as "Mean" or "Total Sample". This provides complete language independence.
  14. Titles: 9 top and 3 bottom titles can be provided.
  15. Additional unlimited number of lines for a table title. These are also used in the optional:
  16. Table of contents.

    Table Layout

  17. The number of rows and columns in a single table is limited only by a maximum of 250,000 cells per table.
  18. A maximum of 91 columns can be printed on a single page. If there are additional columns they are printed on subsequent pages.
  19. The columns of a table are defined by one or more variables which may be nested.
  20. The columns may represent the values of the column (banner) variables, their percents, sums, means or medians.
  21. The rows of a table are defined by a variable or a multiple response group of variables which may be nested.
  22. Rows and columns may be ordered from low to high or high to low.
  23. Rows and columns may be ranked by either ascending or descending frequency.

  24. Ranking may be applied to just some of the values. Controls are for the first N values, or all but the last N values.
  25. Ranking, usually based on the row totals, may instead be based on any column in the table.
  26. The labels file may be used to specify which values to rank.
  27. The labels file may be used to control the exact order of rows or columns.
  28. Many similar tables can be defined with a single statement and processed in the same pass through the data file.
  29. The pieces of the table can be rearranged: For example, the column totals can be at the top, the bottom or even in the middle.
  30. Selected pieces of the table can be ommitted. This permits tables, for example, that have only subtotals, only summary statistics, or have the row totals omitted.

    The Table Body

  31. The table body many contain just the frequencies,
  32. - and/or column percents
  33. - and/or row percents
  34. - and/or percents of total counts
  35. - each of which can be based on the total count,
  36. - and/or the good n (non-missing count),
  37. - or, for multiple response tables, the responses.
  38. Cell expected values and cell chi-square may be requested.
  39. An index value, a measure of how a column compares with the row total may be requested.
  40. When an extra variable is cited, the body may also contain its means,
  41. - and/or its sums,
  42. - and/or its medians.
  43. A choice of formula for computing the medians is available.

    Formatting Options

  44. The cell width can be specified.
  45. The page output width can be specified.
  46. The number of lines per page can be specified.
  47. The width of the margin available for row labels can be specified.
  48. Spacing between the variables in the body can be controlled.
  49. Tabbed output can be requested.
  50. Commas may be properly placed in large numbers.
  51. Empty cells may be left blank or filled with the character of your choice.
  52. Handling of continuation pages includes: continuous printing,
  53. - page break when line count exceeded.
  54. - optional repetition of titles, labels, totals and/or question.
  55. Stub (row) variables can be grouped together on a single page.
  56. Tables can be transposed.
  57. The row totals can be placed on the left or on the right.

    Subtotals and Nets

  58. Subtotals can be defined to add the rows in a table. In a multiple response situation these would be a total of the responses.
  59. Nets can be defined in multiple response tables to total the respondents rather than the responses.
  60. Subtotals and nets can be grouped together or interleaved in the body of the table.
  61. Subtotals and nets usually have the same contents as the cells of the body, but different contents can be requested.
  62. Rows may be ranked within each subtotal/net.
  63. The order of the subtotals/nets can be: - the natural order as determined by the values,
  64. - the order in which the subtotals/nets are defined,
  65. - determined by the frequencies of the subtotals/nets themselves.
  66. Rows with small frequencies can be dynamically combined into an "others" category.

    Summary Statistics

    The following statistics can be included in the summary section. A table can be composed of just the summary statistics.

  67. Total Counts, including missing
  68. Counts on non-missing values
  69. Mean of row (stub) variable or some other variable
  70. Sums of the means variable
  71. Standard deviation of the means variable
  72. Standard error of the means variable
  73. Variance of the means variable
  74. Median of the means variable
  75. Quartiles of the means variable
  76. Low row value
  77. High row value
  78. Range of row values
  79. Mode of row variable
  80. High row value

    Other Statistics

  81. Chi-square with Degrees of Freedom and probability for each 2-way table
  82. Cramer's V for each 2-way table
  83. F-test (or t) and probability each banner variable
  84. Signfificance level for chi-square and F-tests may be specified.
  85. Output for chi-square and F-test may be limited to the probability.
  86. Tests of independence between the cells
  87. Signficance tests of cell differences based on a t-test can be requested for specific variables.
  88. Tests of independence or signficance can be computed between each cell and its row totals.
  89. t-tests between the means of each pair of columns in the summary section can be requested.
  90. All these tests can be done at a specified significance level.

    PostScript and Tabbed Output Formats

  91. Choice of fonts and point sizes for titles, labels, question and the body of the table.
  92. Border around each table.
  93. Optional boxed cells
  94. Control of underlining
  95. Multiple tables per page
  96. Ability to combine tables with text and plots.

    The SURVEY command

  97. Variables may be selected, recoded and generated as the file is input to the SURVEY command.
  98. Multiple input files can be dynamically joined.
  99. Multiple labels files can be supplied.
  100. An optional output file contains the tables in P-STAT system file format.
  101. Include files are supported at both command and subcommand levels.
Return to Market Research Aids &#nbsp; Return to P-STAT's Home Page Return to P-STAT's Home Page









THE EXAMPLES



Reason 1:     Stub and banner variables are usually numeric with value labels used for identification. They can also be character values, as Region and Gender are here. The actual values are used to label the rows and columns.

The Command



SURVEY region;       banner region,     stub gender;

The Table

                      ====== Region ======

                Total
               Sample   East  North  South

 Total              4      1      1      2
 Sample         100.0% 100.0% 100.0% 100.0%

 Gender

 f                  2      1      -      1
                 50.0% 100.0%         50.0%

 m                  2      -      1      1
                 50.0%        100.0%  50.0%

 missing            -      -      -      -

 Base               4      1      1      2
                100.0% 100.0% 100.0% 100.0%
 Mean            1.50   1.00   2.00   1.50
 S.D.            0.58   0.00   0.00   0.71
Return      Return to P-STAT's Home Page


Reason 3:     The Q in ( Q M ) causes banner variable income to be treated as a "quantity" variable. A quantity variable has totals in each cell rather than frequencies. The M requests means as well.

The Command:

SURVEY paceset,     labels 'paceset.lab';
      BAN income ( Q M ) gender,      stub age,
      layout     labels     totals     question     body,     commas;


The Table:

                      ======= Income ====== == Gender ===

                Total
               Sample     Totals       Mean   Male Female

 Total             80  2,476,660  34,398.06     39     40
 Sample         100.0%                       100.0% 100.0%

 Age

 Under 30          28  1,046,870  37,388.21     20      8
                 35.0%                        51.3%  20.0%

 30 to 50          27    703,460  29,310.83     11     16
                 33.8%                        28.2%  40.0%

 Over 50           23    676,020  35,580.00      7     15
                 28.7%                        17.9%  37.5%


Return     Return to P-STAT's Home Page P-STAT Home Page




Reason 5:     MR.STUB identifies a set of variables (VCR to Ans.Mach) that are to be combined as a STUB variable. Here we have 4 yes/no variables and a case can have several yeses, thus a multiple response.

The Command:



SURVEY paceset,     labels 'paceset.lab';
      MR.STUB VCR to Ans.Mach,     ban age gender,
    no missing,     no summary;


The Table:

                           ======== Age ======= == Gender ===

                     Total  Under  30 to   Over
                    Sample     30     50     50   Male Female

 Total Sample           80     28     27     23     39     40
                     100.0% 100.0% 100.0% 100.0% 100.0% 100.0%

 Which of the following do you own?

 Video Cassette         45     19     13     12     27     18
 Recorder             56.2%  67.9%  48.1%  52.2%  69.2%  45.0%

 Compact Disc           28     10     13      5     12     16
 Player               35.0%  35.7%  48.1%  21.7%  30.8%  40.0%

 Portable               35     15      8     10     21     14
 Telephone            43.8%  53.6%  29.6%  43.5%  53.8%  35.0%

 Answering              37     18     12      7     20     16
 Machine              46.2%  64.3%  44.4%  30.4%  51.3%  40.0%


Return     Return to P-STAT's Home Page Return to P-STAT's Home Page




Reason 8:     Value labels and extended variable labels are stored in a separate LABELS file. Here, a 1-78 character extended label for the column variable is printed instead of the 1-16 character variable name.

The Command:



SURVEY election,     LABELS   'election.lab';
      ban q1,     stub gender,     cell.width 9;

The Labels:



q1 'What do you think about the current Republican candidates?'
  (1) do not like (2) no opinion (3) like /
Gender (1) male (2) female /

The Table:


                        ==== What do you think ===
                        ==== about the current ===
                        ====    Republican     ===
                        ====    candidates?    ===

                  Total   do not       no
                 Sample     like  opinion     like

 Total                4        1        2        1
 Sample           100.0%   100.0%   100.0%   100.0%

 Gender

 male                 2        1        1        -
                   50.0%   100.0%    50.0%

 female               2        -        1        1
                   50.0%             50.0%   100.0%

 Missing              -        -        -        -

 Base                 4        1        2        1
                  100.0%   100.0%   100.0%   100.0%
 Mean              1.50     1.00     1.50     2.00
 S.D.              0.58     0.00     0.71     0.00
Return Return     Return to P-STAT's Home Page Return to P-STAT's Home Page


Reason 10:     Labels for the columns can span six lines. The total number of characters that are displayed depends on the cell width. This example shows both an extended label and long (up to 64 characters) value labels for the column variable.

The Command:



SURVEY election,     LABELS 'election.lab';
      ban q1,     stub gender,     CELL.WIDTH 12;

The Labels:



q1 'What do you think about the current Republican candidates?'
  (1) I disagree with their stands on most of the vital issues
  (2) I have given the matter very little thought
  (3) I agree with their positions on the vital issues /

Gender (1) male (2) female /

The Table:

                          === What do you think about the ===
                           ===     current Republican      ===
                           ===         candidates?         ===

                            I disagree                 I agree
                            with their      I have  with their
                             stands on   given the   positions
                           most of the matter very      on the
                                 vital      little       vital
              Total Sample      issues     thought      issues

 Total                   4           1           2           1
 Sample              100.0%      100.0%      100.0%      100.0%

 Gender

 male                    2           1           1           -
                      50.0%      100.0%       50.0%

 female                  2           -           1           1
                      50.0%                   50.0%      100.0%

 Missing                 -           -           -           -

 Base                    4           1           2           1
                     100.0%      100.0%      100.0%      100.0%
 Mean                 1.50        1.00        1.50        2.00
 S.D.                 0.58        0.00        0.71        0.00
Return Return     Return to P-STAT's Home Page Return to P-STAT's Home Page


Reasons 9 and 13:     Variable Gender has 3 lines of extended labels text. When it is used as the stub (row) variable all lines are printed.

The LABELS file has a special variable named SURVEY.LABELS which contains alternate text for programmed text fields. Text field 1 is normally "Total Sample". Here, "Population is printed instead.

The Command:



SURVEY election,     LABELS 'election.lab';
      ban q1,     stub gender,     cell.width 9;

The Labels:



q1 'What do you think about the current Republican candidates?'
    (1) do not like (2) no opinion (3) like /

Gender
    'Gender of the respondent'
    'if Gender could not be determined in a telephone interview'
    'the case was not processed'
    (1) male (2) female /

SURVEY.LABELS (1) Population (2) Good Count (12) Average
  (13) Standard deviation (101) ' ' (m1) No Response /


The Table:

                        ==== What do you think ===
                        ==== about the current ===
                        ====    Republican     ===
                        ====    candidates?    ===

                          do not       no
                            like  opinion     like

 Population           4        1        2        1
                  100.0%   100.0%   100.0%   100.0%

 Gender of the respondent
 if Gender could not be determined in a telephone interview
 the case was not processed

 male                 2        1        1        -
                   50.0%   100.0%    50.0%

 female               2        -        1        1
                   50.0%             50.0%   100.0%

 No Response          -        -        -        -

 Good Count           4        1        2        1
                  100.0%   100.0%   100.0%   100.0%
 Average           1.50     1.00     1.50     2.00
 Standard          0.58     0.00     0.71     0.00
 deviation
Return Return   Return to P-STAT's Home Page Return to P-STAT's Home Page


Reasons 14 and 15:    

In this example the two titles are predefined in a TITLES command. The current date is substituted for .DATE., a P-STAT system variable. LEFT, RIGHT and CENTER can be specified. Local definition of TITLES within SURVEY is also supported.

Table titles may be supplied for each table where a table is defined as a new stub variable or group of stub variables.

The Command:s



TITLES     '.DATE.'     T2     'Pre-election Option Poll' $

SURVEY election,     labels 'election.lab',     TITLES;
    TABLE.TITLES
    'This Survey was conducted in February 1996'
    'By Local Opinion Polls, Inc',
    banner genger,     stub q1,
    margin 20,     no missing,     no summary;

The Table:

              Wed Mar  6  1996              
          Pre-election Option Poll          

 This Survey was conducted in February 1996
 By Local Opinion Polls, Inc

                               == Gender ===

                                 male female

 Population                  4      2      2
                         100.0% 100.0% 100.0%

 What do you think about the current Republican candidates?

 I disagree with             1      1      -
 their stands on most     25.0%  50.0%
 of the vital issues

 I have given the            2      1      1
 matter very little       50.0%  50.0%  50.0%
 thought

 I agree with their          1      -      1
 positions on the         25.0%         50.0%
 vital issues
Return Return     Return to P-STAT's Home Page Return to P-STAT's Home Page


Reason 19:     The keyword WITHIN is used to indicate nesting. Here we have a simple banner variable Own (with an extended label) and then Gender within Own.

The Command:



SURVEY paceset,     labels 'paceset.lab';
      ban own   Gender WITHIN Own,     stub income.groups;

The Table:

                               Respondent
                               ownership         no           yes

                       Total
                      Sample     no    yes   Male Female   Male Female

 Total Sample             80     11     69      5      6     34     34
                       100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0%

 Income of respondent

 Under $20,000             2      1      1      -      1      -      1
                         2.5%   9.1%   1.4%         16.7%          2.9%

 $20,000 to $40,000       50      9     41      4      5     17     23
                        62.5%  81.8%  59.4%  80.0%  83.3%  50.0%  67.6%

 Over $40,000             20      -     20      -      -     10     10
                        25.0%         29.0%                29.4%  29.4%

 Missing                   8      1      7      1      -      7      -
                        10.0%   9.1%  10.1%  20.0%         20.6%

 Base                     72     10     62      4      6     27     34
                        90.0%  90.9%  89.9%  80.0% 100.0%  79.4% 100.0%
 Mean                   2.25   1.90   2.31   2.00   1.83   2.37   2.26
 S.D.                   0.50   0.32   0.50   0.00   0.41   0.49   0.51
Return Return     Return to P-STAT's Home Page Return to P-STAT's Home Page


Reason 20:     Additional columns for each banner variable are requested by key letters in parentheses after the variable name. The Mean, MEDian, Totals and PerCenTs can be requested. The use of "NO V" in the list could be used to remove the individual values.

The Command:



SURVEY paceset,     labels 'paceset.lab';
      BANNER Num.Tv ( M MED ),     stub age;

The Table:

                     ====== Number of television sets in your home: =====

                Total                               5 or
               Sample      1      2      3      4   more     Mean   Median

 Total             80      6     22     18     27      7     3.09     3.00
 Sample         100.0% 100.0% 100.0% 100.0% 100.0% 100.0%

 Age

 Under 30          28      4     14      8      1      1     2.32     2.00
                 35.0%  66.7%  63.6%  44.4%   3.7%  14.3%

 30 to 50          27      1      7      5     13      1     3.22     4.00
                 33.8%  16.7%  31.8%  27.8%  48.1%  14.3%

 Over 50           23      1      1      4     12      5     3.83     4.00
                 28.7%  16.7%   4.5%  22.2%  44.4%  71.4%

 Missing            2      -      -      1      1      -        -        -
                  2.5%                 5.6%   3.7%

 Base              78      6     22     17     26      7     3.08     3.00
                 97.5% 100.0% 100.0%  94.4%  96.3% 100.0%
 Mean            1.94   1.50   1.41   1.76   2.42   2.57
 S.D.            0.81   0.84   0.59   0.83   0.58   0.79
Return Return     Return to P-STAT's Home Page Return to P-STAT's Home Page


Reason 22:     Rows and columns are usually displayed in ascending numeric order. This can by changed by using the ORDER subcommand.

The Command:



SURVEY paceset,     labels 'paceset.lab';
      stub num.tv,     banner age Gender,
      no missing,       no summary,
      ORDER ROWS DOWN;


The Table:

                      ======== Age ======= == Gender ===

                Total  Under  30 to   Over
               Sample     30     50     50   Male Female

 Total             80     28     27     23     39     40
 Sample         100.0% 100.0% 100.0% 100.0% 100.0% 100.0%

 Number of television sets in your home:

 5 or more          7      1      1      5      2      5
                  8.8%   3.6%   3.7%  21.7%   5.1%  12.5%

        4          27      1     13     12      4     22
                 33.8%   3.6%  48.1%  52.2%  10.3%  55.0%

        3          18      8      5      4     11      7
                 22.5%  28.6%  18.5%  17.4%  28.2%  17.5%

        2          22     14      7      1     18      4
                 27.5%  50.0%  25.9%   4.3%  46.2%  10.0%

        1           6      4      1      1      4      2
                  7.5%  14.3%   3.7%   4.3%  10.3%   5.0%
Return Return     Return to P-STAT's Home Page Return to P-STAT's Home Page


Reason 23:     Rows and columns may be ranked by frequency.

The Command:



SURVEY paceset,     labels 'paceset.lab';
      stub num.tv,     banner age gender,
      no missing,       no summary,
      RANK ROWS DOWN;

The Table:

                      ======== Age ======= == Gender ===

                Total  Under  30 to   Over
               Sample     30     50     50   Male Female

 Total             80     28     27     23     39     40
 Sample         100.0% 100.0% 100.0% 100.0% 100.0% 100.0%

 Number of television sets in your home:

        4          27      1     13     12      4     22
                 33.8%   3.6%  48.1%  52.2%  10.3%  55.0%

        2          22     14      7      1     18      4
                 27.5%  50.0%  25.9%   4.3%  46.2%  10.0%

        3          18      8      5      4     11      7
                 22.5%  28.6%  18.5%  17.4%  28.2%  17.5%

 5 or more          7      1      1      5      2      5
                  8.8%   3.6%   3.7%  21.7%   5.1%  12.5%

        1           6      4      1      1      4      2
                  7.5%  14.3%   3.7%   4.3%  10.3%   5.0%
Return Return     Return to P-STAT's Home Page Return to P-STAT's Home Page


Reason 24:     Ranking can be controlled so only some of the values are ranked. If the number is positive, only the first N values are ranked. If the number is negative, all but the last N values are ranked.

The Command:



SURVEY paceset,     labels 'paceset.lab';
      stub num.tv,     banner age gender,
      RANK ROWS -2 DOWN,
      no missing,     no summary;

The Table:

                      ======== Age ======= == Gender ===

                Total  Under  30 to   Over
               Sample     30     50     50   Male Female

 Total             80     28     27     23     39     40
 Sample         100.0% 100.0% 100.0% 100.0% 100.0% 100.0%

 Number of television sets in your home:

        2          22     14      7      1     18      4
                 27.5%  50.0%  25.9%   4.3%  46.2%  10.0%

        3          18      8      5      4     11      7
                 22.5%  28.6%  18.5%  17.4%  28.2%  17.5%

        1           6      4      1      1      4      2
                  7.5%  14.3%   3.7%   4.3%  10.3%   5.0%

        4          27      1     13     12      4     22
                 33.8%   3.6%  48.1%  52.2%  10.3%  55.0%

 5 or more          7      1      1      5      2      5
                  8.8%   3.6%   3.7%  21.7%   5.1%  12.5%
Return     Return to P-STAT's Home Page


Reason 28:     Multiple STUB, GR.STUB and MR.STUB subcommands, with optional table titles, can be processed in a single pass of the file when the banner variables and the table options remain the smae.

The Command:



SURVEY s0296,   labels 's0296.lab';

    layout     question     labels     totals.area     body     summary,
    banner Age Gender Occupation Education,

    STUB           Q1A to Q39,
    MR.STUB     Q40A to Q40G,
    STUB           Q43, Q45, Q90 TO 103
    TABLE.TITLES 'These two tables are to be grouped together',
    GR.STUB     Q42 Q44 ;


    TABLE.TITLES 'These tables have means on variable income'
    STUB Item33
   
TABLES.TITLES 'Item 33 will be processed in the same pass as'
    STUB Item35
    TABLE.TITLES 'Item 35 and Q244 though each as a different table title'
    Q244,
    MEANS Income;

Return     Return to P-STAT's Home Page




Reason 29:     The LAYOUT subcommand can be used to rearrange pieces of a table. Any piece that is not mentioned is omitted.

The Command:



SURVEY paceset,     labels 'paceset.lab';
      stub own,     banner age gender,

      LAYOUT QUESTION LABELS BODY TOTALS;

The Table:

 Respondent ownership of electronic items

                      ======== Age ======= == Gender ===

                Total  Under  30 to   Over
               Sample     30     50     50   Male Female

 no                11      2      6      3      5      6
                 13.8%   7.1%  22.2%  13.0%  12.8%  15.0%

 yes               69     26     21     20     34     34
                 86.2%  92.9%  77.8%  87.0%  87.2%  85.0%

 Total             80     28     27     23     39     40
 Sample         100.0% 100.0% 100.0% 100.0% 100.0% 100.0%


Return     Return to P-STAT's Home Page


Reason 30:     The row totals which are usually printed as the first column of a table can be omitted or moved to the right.

The Command:



SURVEY paceset,     labels 'paceset.lab';
      stub own,     banner age gender,
      layout     question     labels     body     totals,

      NO ROW.TOTALS;

The Table:

 Respondent ownership of electronic items

               ======== Age ======= == Gender ===

                Under  30 to   Over
                   30     50     50   Male Female

 no                 2      6      3      5      6
                  7.1%  22.2%  13.0%  12.8%  15.0%

 yes               26     21     20     34     34
                 92.9%  77.8%  87.0%  87.2%  85.0%
 
 Total             28     27     23     39     40
 Sample         100.0% 100.0% 100.0% 100.0% 100.0%


Return     Return to P-STAT's Home Page


Reason 31:     The body of the table usually contains both the counts (frequencies) and the column percents. However the contents can be controlled with the BODY CONTAINS subcommand.    

The Command:



SURVEY paceset,     labels 'paceset.lab';
      stub own,     banner age gender,
      layout     question     labels     body,

      BODY     CONTAINS     COUNTS;

The Table:

 Respondent ownership of electronic items

                      ======== Age ======= == Gender ===

                Total  Under  30 to   Over
               Sample     30     50     50   Male Female

 no                11      2      6      3      5      6

 yes               69     26     21     20     34     34
 

Return     Return to P-STAT's Home Page


Reason 33:     A combination of the BODY subcommand and the PERCENTS subcommand is used to get a table in which the cells contains just the row and column percents. Note: The percents are printed in the order specified.

The Command:



SURVEY paceset,     labels 'paceset.lab';
      stub own,     banner age gender,
      layout     question     labels     body,

      ROW     COLUMN     PERCENTS,
      BODY     CONTAINS     PERCENTS;


The Table:

 Respondent ownership of electronic items

                      ======== Age ======= == Gender ===

                Total  Under  30 to   Over
               Sample     30     50     50   Male Female

 no             100.0%  18.2%  54.5%  27.3%  45.5%  54.5%
                 13.8%   7.1%  22.2%  13.0%  12.8%  15.0%

 yes            100.0%  37.7%  30.4%  29.0%  49.3%  49.3%
                 86.2%  92.9%  77.8%  87.0%  87.2%  85.0%

Return     Return to P-STAT's Home Page


Reason 36:     Percents are usually based on the total n (count). The base can be changed to the good n (missing is excluded) or, in a multiple response situation to the responses. This example has row percents based on both the total n and the good n.

The Command:



SURVEY paceset,     labels 'paceset.lab';
      stub income.groups,     banner age ( T ) gender ( T ),
      layout     labels     question     body,
      margin 10,

      ROW PERCENTS     BASED     GOOD.N,
      ROW PERCENTS     BASED     TOTAL.N;


The Table:

                     =========== Age =========== ====== Gender ======

               Total  Under  30 to   Over
              Sample     30     50     50 Good N   Male Female Good N

 Income of respondent

 Under             2      -      2      -      2      -      2      2
 $20,000       100.0%        100.0%                      100.0%
               100.0%        100.0%                      100.0%

 $20,000 to       50     18     19     13     50     21     28     49
 $40,000       100.0%  36.0%  38.0%  26.0%         42.9%  57.1%
               100.0%  36.0%  38.0%  26.0%         42.0%  56.0%

 Over             20     10      3      6     19     10     10     20
 $40,000       100.0%  52.6%  15.8%  31.6%         50.0%  50.0%
               100.0%  50.0%  15.0%  30.0%         50.0%  50.0%
Return     Return to P-STAT's Home Page


Reason 38:     The contribution of each cell to a possible chi square statistic is requested with the CELL.CHI subcommand. The number of cases that are expected in the cell given the marginals is requested with the EXPECTED.VALUES subcommand.

The Command:



SURVEY paceset,     labels 'paceset.lab';

      stub income.groups,     banner age gender,
      layout     labels     totals     question     body,
      margin 10,     no percents,

      CELL.CHI,     EXPECTED.VALUE;

The Table:

                     ======== Age ======= == Gender ===

               Total  Under  30 to   Over
              Sample     30     50     50   Male Female

 Total
 Sample           80     28     27     23     39     40

 Income of respondent

 Under             2      -      2      -      -      2
 $20,000                0.8    0.7    0.5    0.9    1.1
                        0.8    2.6    0.5    0.9    0.7

 $20,000 to       50     18     19     13     21     28
 $40,000               19.7   16.9   13.4   21.4   27.6
                        0.1    0.3    0.0    0.0    0.0

 Over             20     10      3      6     10     10
 $40,000                7.5    6.4    5.1    8.7   11.3
                        0.8    1.8    0.2    0.2    0.1
Return     Return to P-STAT's Home Page


Reason 40:     Means on another variable can be requested for the body of the table as well as the summary.

The Command:



SURVEY paceset, labels 'paceset.lab';

      banner age gender,     stub income.groups,     MEANS Income,
      BODY     CONTAINS     COUNTS     MEANS,
      SUMMARY       CONTAINS     BASE     MEANS,

      no missing,     COMMAS,     places means 0;

The Table:

                             ========= Age ========= === Gender ====

                       Total   Under   30 to
                      Sample      30      50 Over 50    Male  Female

 Total Sample             72      28      24      19      31      40
                       100.0%  100.0%  100.0%  100.0%  100.0%  100.0%

 Income of respondent

 Under $20,000             2       -       2       -       -       2
                      18,400          18,400                  18,400

 $20,000 to $40,000       50      18      19      13      21      28
                      30,007  31,359  28,224  30,740  31,291  28,980

 Over $40,000             20      10       3       6      10      10
                      46,976  48,240  43,467  46,067  48,240  45,711

 Base                     72      28      24      19      31      40
                       100.0%  100.0%  100.0%  100.0%  100.0%  100.0%
 Mean Income          34,398  37,388  29,311  35,580  36,759  32,634
Return     Return to P-STAT's Home Page


Reason 42:     Medians on an external variable ban be computed for the body as well as the summary section. Means and medians in the same table must be on the same variable.

The Command:



SURVEY paceset, labels 'paceset.lab';

      no missing,     commas,     places means 0,     places medians 0,
      table.title 'Cells contain counts' 'means of income' 'medians of Income',
      banner age gender,     stub income.groups,     MEANS Income,

      BODY     CONTAINS     COUNTS     MEANS     MEDIANS,
      SUMMARY     CONTAINS     MEANS     MEDIANS;


The Table:

 Cells contain counts
 means of income
 medians of Income

                             ========= Age ========= === Gender ====

                       Total   Under   30 to
                      Sample      30      50 Over 50    Male  Female

 Total Sample             72      28      24      19      31      40
                       100.0%  100.0%  100.0%  100.0%  100.0%  100.0%

 Income of respondent

 Under $20,000             2       -       2       -       -       2
                      18,400          18,400                  18,400
                      18,400          18,400                  18,400

 $20,000 to $40,000       50      18      19      13      21      28
                      30,007  31,359  28,224  30,740  31,291  28,980
                      30,085  31,350  25,900  31,800  32,500  27,650

 Over $40,000             20      10       3       6      10      10
                      46,976  48,240  43,467  46,067  48,240  45,711
                      45,850  46,950  42,500  46,050  46,950  45,400

 Mean Income          34,398  37,388  29,311  35,580  36,759  32,634
 Median               32,510  34,550  25,950  32,520  34,700  30,900
Return     Return to P-STAT's Home Page


Reason 51:     Cells with no contents are normally represented with a dash on the first row in the cell. This can be changed to any character of your choice with the CHARACTER subcommand. When two characters are supplied, all rows in the cell are filled.

The Command:



SURVEY paceset,     labels 'paceset.lab';

      banner age gender,     stub income.groups,
      CHARACTER ZERO     '0' '0';

The Table:

                             ======== Age ======= == Gender ===

                       Total  Under  30 to   Over
                      Sample     30     50     50   Male Female

 Total Sample             80     28     27     23     39     40
                       100.0% 100.0% 100.0% 100.0% 100.0% 100.0%

 Income of respondent

 Under $20,000             2      0      2      0      0      2
                         2.5%   0.0%   7.4%   0.0%   0.0%   5.0%

 $20,000 to $40,000       50     18     19     13     21     28
                        62.5%  64.3%  70.4%  56.5%  53.8%  70.0%

 Over $40,000             20     10      3      6     10     10
                        25.0%  35.7%  11.1%  26.1%  25.6%  25.0%

 Missing                   8      0      3      4      8      0
                        10.0%   0.0%  11.1%  17.4%  20.5%   0.0%

 Base                     72     28     24     19     31     40
                        90.0% 100.0%  88.9%  82.6%  79.5% 100.0%
 Mean                   2.25   2.36   2.04   2.32   2.32   2.20
 S.D.                   0.50   0.49   0.46   0.48   0.48   0.52
Return     Return to P-STAT's Home Page


Reason 55:     A group of variables can be printed on a single page and share the labels and totals areas of the display. This is often used for tables of summary statistics.

The Command:



SURVEY paceset,     labels 'paceset.lab';
      banner num.tv,     GR.STUB     Age Gender Income.Groups,
      layout     labels     question     summary,     no base,     margin 10;

The Table:

                     == Number of television sets in ==
                     ==          your home:          ==

               Total                               5 or
              Sample      1      2      3      4   more

 Age

 Mean           1.94   1.50   1.41   1.76   2.42   2.57
 S.D.           0.81   0.84   0.59   0.83   0.58   0.79

 Gender

 Mean           1.51   1.33   1.18   1.39   1.85   1.71
 S.D.           0.50   0.52   0.39   0.50   0.37   0.49

 Income of respondent

 Mean           2.25   2.17   2.29   2.17   2.23   2.43
 S.D.           0.50   0.75   0.46   0.39   0.51   0.53
Return     Return to P-STAT's Home Page


Reason 56:     Tables can be transposed so that the banner variable defines the rows and the stub variable defines the columns. The layout now has the summary section as columns rather than rows. This is often used with GR.STUBS (grouped variables) to produce tables of marginals.

The Command:



SURVEY paceset;
      gr.stub     age gender income.groups,
      margin 10,     TRANSPOSE;

The Table:

              Total                        Miss
             Sample      1      2      3    ing    Base    Mean    S.D.

 Age             80     28     27     23      2      78    1.94    0.81
              100.0%  35.0%  33.8%  28.7%   2.5%   97.5%

 Gender          80     39     40      -      1      79    1.51    0.50
              100.0%  48.8%  50.0%          1.2%   98.8%

 Income.         80      2     50     20      8      72    2.25    0.50
 Groups       100.0%   2.5%  62.5%  25.0%  10.0%   90.0%

Return     Return to P-STAT's Home Page


Reason 58:     Subtotals are automatically included in the table when DEFINE statements provide the labels, variables and values to be totaled.

The Command:



SURVEY paceset, labels 'paceset.lab';    
      DEFINE 'Some Television Sets' NUM.TV 1 to 3,
      DEFINE 'Many Television Sets' NUM.TV 4 to 6,

      stub num.tv,     banner age gender,     no missing,     no summary;

The Table:

                Total  Under  30 to   Over
               Sample     30     50     50   Male Female

 Total             80     28     27     23     39     40
 Sample         100.0% 100.0% 100.0% 100.0% 100.0% 100.0%

 Number of television sets in your home:

        1           6      4      1      1      4      2
                  7.5%  14.3%   3.7%   4.3%  10.3%   5.0%

        2          22     14      7      1     18      4
                 27.5%  50.0%  25.9%   4.3%  46.2%  10.0%

        3          18      8      5      4     11      7
                 22.5%  28.6%  18.5%  17.4%  28.2%  17.5%

        4          27      1     13     12      4     22
                 33.8%   3.6%  48.1%  52.2%  10.3%  55.0%

 5 or more          7      1      1      5      2      5
                  8.8%   3.6%   3.7%  21.7%   5.1%  12.5%

 Some              46     26     13      6     33     13
 Television      57.5%  92.9%  48.1%  26.1%  84.6%  32.5%
 Sets
 ----------

 Many              34      2     14     17      6     27
 Television      42.5%   7.1%  51.9%  73.9%  15.4%  67.5%
 Sets
 ----------
Return     Return to P-STAT's Home Page


Reason 60:     Subtotals and nets can be inserted into the body of the table. They can either precede or follow the values included in their definitions.

The Command:



SURVEY paceset,     labels 'paceset.lab';
      define 'Some Television Sets' NUM.TV 1 to 3,
      define 'Many Television Sets' NUM.TV 4 to 6,
      stub num.tv,     banner age gender,     no missing,     no summary,

      SUBTOTALS     BEFORE     LOW VALUES;

The Table:

                      ======== Age ======= == Gender ===

                Total  Under  30 to   Over
               Sample     30     50     50   Male Female

 Total             80     28     27     23     39     40
 Sample         100.0% 100.0% 100.0% 100.0% 100.0% 100.0%

 Number of television sets in your home:

 Some              46     26     13      6     33     13
 Television      57.5%  92.9%  48.1%  26.1%  84.6%  32.5%
 Sets
 ----------

        1           6      4      1      1      4      2
                  7.5%  14.3%   3.7%   4.3%  10.3%   5.0%

        2          22     14      7      1     18      4
                 27.5%  50.0%  25.9%   4.3%  46.2%  10.0%

        3          18      8      5      4     11      7
                 22.5%  28.6%  18.5%  17.4%  28.2%  17.5%

 Many              34      2     14     17      6     27
 Television      42.5%   7.1%  51.9%  73.9%  15.4%  67.5%
 Sets
 ----------

        4          27      1     13     12      4     22
                 33.8%   3.6%  48.1%  52.2%  10.3%  55.0%

 5 or more          7      1      1      5      2      5
                  8.8%   3.6%   3.7%  21.7%   5.1%  12.5%
Return     Return to P-STAT's Home Page


Reason 64:     When rank is used with subtotals, the order of the subtotals can be specified.

The Command:



SURVEY pets,     labels `pets.lab';

      define `birds' pet 1 6,
      define `mammals' pet 2 3 5,
      define `other animals' pet 4 7 8,

      banner gender age, stub pet,     no percents,     no missing,     no summary,
      margin 16,     gap 2,     no skip,     indent 4,
      subtotals     before     low value,
      RANK     WITHIN     SUBTOTALS;
      $


The Table:

                             == Gender ===   ======== Age =======

                     Total                    Under  20 to 40 and
                    Sample     male female       20     39   over

 Total Sample          925      478    447      261    319    345

 What animal makes the best pet?

 birds                 155      100     55       28     48     79
 -----
     parakeet           80       25     55       16     25     39
     canary             75       75      -       12     23     40
 mammals               675      325    350      220    230    225
 -------
     cat               350      150    200      100    110    140
     dog               275      150    125      100    100     75
     horse              50       25     25       20     20     10
 other animals          95       53     42       13     41     41
 -------------
     fish               90       49     41        9     40     41
     turtle              4        3      1        3      1      -
     snake               1        1      -        1      -      -
Return     Return to P-STAT's Home Page


Reason 65:     When rank is used with subtotals, you can request that the subtotals be ranked and the rows ranked within them.

The Command:



SURVEY pets,     labels `pets.lab';

      define `birds' pet 1 6,
      define `mammals' pet 2 3 5,
      define `other animals' pet 4 7 8,
     
      banner gender age, stub pet,     no percents,     no missing,     no summary,
      margin 16,     gap 2,     no skip,     indent 4,
      subtotals     before     low value,
      RANK CONTROLS SUBTOTALS;
      $


The Table:

                             == Gender ===   ======== age ======= 

Total Under 20 to 40 and
Sample male female 20 39 over

Total Sample 925 478 447 261 319 345

What animal makes the best pet?

mammals 675 325 350 220 230 225
-------
cat 350 150 200 100 110 140
dog 275 150 125 100 100 75
horse 50 25 25 20 20 10
birds 155 100 55 28 48 79
-----
parakeet 80 25 55 16 25 39
canary 75 75 - 12 23 40
other animals 95 53 42 13 41 41
-------------
fish 90 49 41 9 40 41
turtle 4 3 1 3 1 -
snake 1 1 - 1 - -
Return     Return to P-STAT's Home Page


Reason 67:     The summary section usually includes the base (good n), means and standard deviation. This can be expanded to include a number of other univariate statistics.

The Command:



SURVEY paceset,     labels `paceset.lab' `survey.lab',     weight weight;

      layout     labels     summary,     banner age own,     stub income.groups,
      margin 20,     cell.width 8,     gap 2,     no percents,

      SUMMARY     CONTAINS     TOTAL.N     UNWEIGHTED.N
      BASE     UNWEIGHTED.BASE     SUM     MEAN
      S.D.     VARIANCE     STANDARD.ERROR,
      MODE   QUARTILES     LOW     HIGH     RANGE;


File survey.lab contains:

Weight ( xl ` ' ) /
SURVEY.LABELS 
(1) Population:Weighted (2) Good N:Weighted 
(4) Unweighted total    (5) Unweighted Good N 
(11) Sums               (12) Average 
(13) Standard deviation (14) S.D. squared  (16) Q2 
(18) Low value          (19) High value    (20) High - Low   
(21) Q1                 (22) Q3           (101) Respon*dents 
(202) ` '               (211) No Reply to the question / 


The Table:

                                                        == Respondent =
                              ========= Age =========   == ownership  =

                     Respon     Under   30 to
                      dents        30      50 Over 50        no     yes

 Population:             82        25      26      29        11      71
 Weighted
 Unweighted total        80        28      27      23        11      69
 Good N:Weighted         73        25      23      24        10      63
 Unweighted Good N       72        28      24      19        10      62
 Sums                   164        59      48      55        19     145
 Average               2.25      2.36    2.04    2.32      1.91    2.31
 Standard deviation    0.50      0.49    0.46    0.48      0.31    0.50
 S.D. squared          0.25      0.24    0.21    0.23      0.10    0.25
 Standard error        0.06      0.10    0.10    0.10      0.10    0.06
 Mode                     2         2       2       2         2       2
 Q1                    2.00      2.00    2.00    2.00      2.00    2.00
 Q2                    2.00      2.00    2.00    2.00      2.00    2.00
 Q3                    3.00      3.00    2.00    3.00      2.00    3.00
 High value               1         2       1       2         1       1
 Low value                3         3       3       3         2       3
 High - Low               2         1       2       1         1       2
Return     Return to P-STAT's Home Page


Reason 81:     The chi-square can be requested for each 2-way table.

The Command:



SURVEY paceset,     LABELS `paceset.lab' ;
      banner income.group,    & stub age,
      no missing,    & skip 0,
      CHI;

The Table:

                                 $20,
                         Under 000 to
                  Total   $20,   $40, Over $
                 Sample    000    000 40,000

   Total             80      2     50     20
   Sample         100.0% 100.0% 100.0% 100.0%

   Age

   Under 30          28      -     18     10
                   35.0%         36.0%  50.0%
   30 to 50          27      2     19      3
                   33.8% 100.0%  38.0%  15.0%
   Over 50           23      -     13      6
                   28.7%         26.0%  30.0%

   Base              78      2     50     19
                   97.5% 100.0% 100.0%  95.0%
   Mean            1.94   2.00   1.90   1.79
   S.D.            0.81   0.00   0.79   0.92

   Chi Square             7.1653 *
   DF for Chi             4.
   Probability            0.1274
   * Some cells had an expected value of less than 5.

Return     Return to P-STAT's Home Page


Reason 83:     F tests can be requested. If the column variable has only 2 values, the t value is also reported.

The Command:



SURVEY cell,     labels `cell.lab';
      banner process treatment,     group.stubs Efficiency Rating,
      places means 3,     cell.width 9,     margin 14,
      layout     labels     question     summary,
      summary     contains     means     s.d     s.e,     F.TEST;

The Table:

                                                 ===   Bake    ===
                      ===== Processing Lab ===== === Treatment ===

                Total
               Sample       TG       SW       BK    Short     Long

 Efficiency

 Mean           4.921    5.470    4.066    5.276    4.175    5.621
 S.D.           1.103    0.490    1.413    0.496    1.151    0.326
 Standard error 0.138    0.110    0.301    0.106    0.207    0.057

 F  value                 14.7035                    47.9747
 Probability               0.0000                     0.0000
 t  value                                             6.9264

 Rating

 Mean          18.764   13.959   27.103   14.794   24.024   13.823
 S.D.          12.466    4.978   17.471    5.213   15.823    4.387
 Standard error 1.558    1.113    3.725    1.112    2.842    0.764

 F  value                  9.5711                    12.6898
 Probability               0.0002                     0.0007
 t  value                                             3.5623
Return     Return to P-STAT's Home Page


Reason 87:     Significance tests are used to determine if there are differences between cells that should be investigated further.

The Command:



SURVEY t3,      labels `test.lab';
      banner region,      stub gender,
      no missing,     summary     contains     means,
      TEST      PROPORTION     .90;

The Table:

                     ====== Region ======

               Total  North  South
              Sample   East   East   West
                          A      B      C

Total            990    191    357    442
Sample         100.0% 100.0% 100.0% 100.0%

Gender

female           182     51     65     66
                18.4%  26.7%  18.2%  14.9%
                         BC

male             808    140    292    376
                81.6%  73.3%  81.8%  85.1%
                                 A      A

Mean            1.82   1.73   1.82   1.85

.90 significance level used for t-test of proportions



Return     Return to P-STAT's Home Page


Reason 89:     t-tests between each pair of means for each of the banner variables can be requested.

The Command:



SURVEY cell,     labels `cell.lab';

      banner process treatment,     group.stubs efficiency rating,
      places means 3,      cell.width 9,      margin 14,
      layout      labels      question      summary,
      summary     contains     means     s.d     s.e,     f.test,
      TEST MEANS;

The Table:

                                                      ===   Bake    ===
                           ===== Processing Lab ===== === Treatment ===

                     Total
                    Sample       TG       SW       BK    Short     Long
                                  A        B        C        D        E

 Efficiency

 S.D.                1.103    0.490    1.413    0.496    1.151    0.326
 Standard error      0.138    0.110    0.301    0.106    0.207    0.057
 Mean                4.921    5.470    4.066    5.276    4.175    5.621
                                  B                 B                 D

 F  value                      14.7035                    47.9747
 Probability                    0.0000                     0.0000
 t  value                                                  6.9264

 Rating

 S.D.               12.466    4.978   17.471    5.213   15.823    4.387
 Standard error      1.558    1.113    3.725    1.112    2.842    0.764
 Mean               18.764   13.959   27.103   14.794   24.024   13.823
                                          AC                 E

 F  value                       9.5711                    12.6898
 Probability                    0.0002                     0.0007
 t  value                                                  3.5623

 .95 significance level used for t-test of means
Return     Return to P-STAT's Home Page


Reason 91:     PostScript output can be requested. Different fonts can be specified for titles, labels, the question itself and the body of the table. The orientation can be changed and multiple surveys can be placed on a single page.

The Command:



SURVEY paceSet,     labels `paceset.lab',
      pr `paceSet.ps',      titles,
      POSTSCRIPT,      PORTRAIT,      FORMAT 2 ,
      FONT 'ZapfChancery-MediumItalic' 10;

      banner age gender,     stub income.groups,     mean income,
      places means 0,       cell.width 9,       char for percent ' ' ;     

The Table: