Shareware Grab Bag

home *** CD-ROM | disk | FTP | other *** search

/ Shareware Grab Bag / Shareware Grab Bag.iso / 090 / sampler4.arc / CROSSTAB.SAC < prev next >

Wrap

Text File | 1985-08-30 | 13.4 KB | 385 lines

ON ECHO !*******************************************************************! ! ! ! Cross-section Techniques in SORITEC Sampler ! ! (Chapter 8) ! ! ! !*******************************************************************! ! ! SORITEC Sampler provides two commands for analyzing cross-section ! data. ! ! SYNOPSIS produces a detailed summary analysis of data series ! including mean, standard deviation, median, quartiles, deciles, ! variance, skewness, kurtosis, coefficient of variation ! number of observations, number of missing values, minimum, ! range, mode and frequency of the mode. ! ! XTAB calculates a standard row * column crosstabulation report. ! Crosstabulation output can be controlled by the interactive ! print server if the ON CRT option is enabled. ! ! These commands are demonstrated using sample data drawn from ! Berenson, M.L., D.M. Levine and M. Goldstein, Intermediate ! Statistical Methods and Applications: A Computer Package ! Approach. Englewood Cliffs: Prentice-Hall, 1983, p. 30. The ! data represent raw data derived from an hypothetical employee ! questionnaire and were used to demonstrate features of various ! commercial statistical packages. The data base contains the ! sex, age, education, employment tenure, job class and weekly ! salary of 46 respondents. ! ! Read in the sample data. ! USE 1 46 READ sex age education months_employed job_classification salary 1 40 3 207 2 458 0 21 2 25 1 235 1 49 2 390 3 798 1 45 3 34 2 339 1 25 5 20 2 339 1 47 4 209 3 584 1 22 3 15 2 296 1 22 3 12 2 235 1 28 4 70 2 235 1 57 3 475 3 571 0 20 2 31 2 283 1 50 3 364 2 403 0 51 3 31 1 363 1 62 2 416 2 436 1 32 2 129 2 435 1 51 4 42 1 334 1 55 4 89 2 455 0 45 2 274 1 293 1 31 5 15 1 228 1 62 2 372 2 379 1 54 2 83 2 339 0 42 2 98 2 270 1 26 4 4 2 303 1 38 4 43 2 340 1 32 2 85 2 359 1 45 4 26 2 334 1 28 3 99 2 314 1 42 4 8 2 340 1 50 2 82 2 373 1 44 4 211 3 581 1 35 4 137 2 451 1 68 2 128 2 323 1 40 5 18 2 345 1 24 5 5 2 333 0 27 4 21 1 296 1 46 5 87 2 363 0 63 3 102 2 325 1 28 5 60 2 363 1 64 1 338 2 323 1 51 2 351 2 323 1 47 1 101 2 323 0 22 3 28 1 256 1 51 5 317 3 515 1 29 5 31 3 411 1 48 4 316 3 544 1 59 5 293 3 450 ; END cls !*******************************************************************! ! ! ! Summary Statistics of Cross-section Data ! ! (Section 8.1) ! ! ! !*******************************************************************! ! ! A summary analysis of any variable is produced by the SYNOPSIS ! command, i.e., ! SYNOPSIS age months_employed salary ! ! All summary statistics produced at the terminal are stored as ! internal SORITEC variables and can be retrieved explicitly by ! the RECOVER command or referenced implicitly in command lines ! by prefixing the internal variable name with an up-carat(^). ! Internal variable names are stored as vectors with the same ! number of elements as arguments in the command line. For example, ! recovering the means of the variables is accomplished with the ! command: ! RECOVER mean_values MEANS ! ! Print out the vector. ! PRINT mean_values ! ! Now if we want to access the mean of "months_employed" ... ! SET mean_months_employed = mean_values(2) PRINT mean_months_employed ! ! As another example, recover the vector of minimum values ! RECOVER minimum_values MIN ! ! Currently, deciles and quartiles are stored as vectors, meaning ! that SORITEC Sampler only retains decile and quartile values of ! the last variable in the argument list. Therefore, after ! executing the command: ! RECOVER salary_quartiles QUARTIL ! ! "quartile_values" contains statistics on the variable "salary". ! PRINT salary_quartiles ! ! Note that the same results would be produced by the command: ! PRINT ^QUARTIL ! ! We'll recover the quartile values for "months_employed" for the ! crosstabulation analysis below. To retrieve the data, we must ! re-execute the SYNOPSIS command with "months_employed" as the ! last argument. ! SYNOPSIS months_employed RECOVER months_employed_quartiles QUARTIL ! ! That's about it for the SYNOPSIS command. ! cls !*******************************************************************! ! ! ! Crosstabulation Analysis ! ! (Section 8.2) ! ! ! !*******************************************************************! ! ! Crosstabulation tables are generated by the XTAB command in ! SORITEC Sampler. Only a two-way table may be generated, meaning ! that only two arguments may follow the command name. The ! data series must be DISCRETE data. We'll show you later in this ! demo how to use the RECODE command to transform continuous to ! discrete data. ! ! Suppose we want to crosstabulate job classification with education. ! Both of these variables consist of discrete data with the following ! meanings: ! ! Job Classification: (1) clerical ! (2) technical ! (3) managerial ! ! Education: (1) non-high school graduate ! (2) high school graduate ! (3) post-high school education ! (4) 2-year college degree ! (5) 4-year college degree ! ! A crosstabulated table is produced by the command: ! XTAB job_classification education ! ! Note that the first argument is the row variable and the second is the ! column variable. ! ! Recoverable internal SORITEC variables are: ! NROW - the number of rows in the table ! NCOL - number of columns ! RMARGIN - vector containing row margin values ! CMARGIN - vector containing column values ! XTABLE - matrix of the inner table ! ! Note that the inner table is stored only when the NOMATS option ! is OFF (the default). ! ! The inner table can be printed directly by the command. ! PRINT ^XTABLE ! ! When data series are continuous, they must first be transformed into ! discrete values before they can be crosstabulated. This is done ! in SORITEC Sampler with the RECODE command (Section 6.3). ! With the RECODE command, the arguments in the command line are ! the output data series, the input data series and the interval ! boundaries over which you want the data categorized, ranging from ! minimum to maximum values. Interval boundaries may be numbers, ! CONSTANTS or PARAMETERS. ! ! For example, suppose we want to crosstabulate sex versus salary, ! and months employed versus salary. We'll use the quartiles and ! minimums we saved from the SYNOPSIS command to define the quantiles ! and automate the procedure as much as possible. ! ! The first crosstabulation will be sex versus salary. ! ! First, recover the current USE period and store it for later use ! in the variable "use_period" ! RECOVER use_period USE ! ! Now define the quantiles, which we will call p1, p2, p3, p4 and p5 ! as the GROUP "pgroup". ! GROUP pgroup p1 p2 p3 p4 p5 ! ! Assign the minimum interval boundary, p1, to be the minimum salary ! of the sample, which we saved in the VECTOR "minimum_values". ! The minimum salary is the 3rd element of this vector. ! SET p1 = minimum_values(3) ! ! Since the interval boundaries in the RECODE command must be ! numbers, CONSTANTS or PARAMETERS, we can either extract individual ! elements from the vector of quartiles we will use for the other ! interval boundaries or explicitly reference each vector element ! in the RECODE argument list. In the first crosstabulation, we'll ! extract the individual elements of the vector and assign them to ! CONSTANTS to demonstrate some of the tricks you can use when ! applying SORITEC Sampler. In the second example, we'll take a ! "short-cut" approach. ! ! In SORITEC, extraction of individual elements from data series ! is facilitated by the GETOBS command. This command is not ! supported by SORITEC Sampler, however. We can still extract ! individual elements and assign them to CONSTANTS with the ! following set of commands. ! ! First, change the USE period to the length of the vector of quartiles. ! USE 1 4 ! ! Then transform the vector into a data series. ! temp = salary_quartiles ! ! Lastly, set up a DOT-loop that sequentially assigns each element ! of the data series to the remaining interval boundaries for the ! RECODE command. ! ! ! Initialize an index counter for accessing each element of the ! temporary data series, "temp" ! SET j = 0 DOT p2 p3 p4 p5 SET j = j + 1 ! ! Reset the USE period to each element of the data series. ! USE j ! ! Assign an element of the data series to an interval boundary. ! SET : = temp ENDDOT ! ! Print out the group to insure the correct values have been assinged ! to the interval boundaries. (Refer to the interval boundaries as a ! group.) ! ON GROUP PRINT pgroup ! ! Now restore the original USE period which we previously recovered. ! USE use_period ! ! Check to be sure the original USE period has been restored. ! USE ! And use the RECODE command to transform the continuous data series, ! "salary" to the discrete series "salary_levels". ! RECODE salary_levels salary pgroup ! ! Print out both variables. ! PRINT salary salary_levels ! ! Finally, crosstabulate sex versus salary_levels. ! Sex is defined 0 = female and 1 = male in the data. ! XTAB sex salary_levels ! ! Repeat the procedure for months employed versus salary. ! This time, however, reference the relevant elements of the ! vectors "Minimum_values" and "months_employed_quartiles" ! directly into the RECODE command line. ! RECODE months_employed_intervals months_employed minimum_values(2) & months_employed_quartiles(1) months_employed_quartiles(2) & months_employed_quartiles(3) months_employed_quartiles(4) ! ! Not as elegant, but it does the trick! See? ! PRINT months_employed months_employed_intervals ! ! Now generate the crosstabulation table. ! XTAB months_employed_intervals salary_levels ! ! That's it! ! cls QUIT