POLS 6386 MEASUREMENT THEORY
First Assignment
Due 7 February 2001


  1. The aim of this problem is to familarize you with the classic KYST scaling program. Download the program

    KYSTBIG.FOR -- Program can be compiled by using gfortran

    and the sample data file

    KYST Supreme Court Data

    and place them in the same folder on a WINTEL machine.

    The sample data file is reproduced below. It contains the lower half of an agreement score matrix computed between the 31 supreme court justices who served on the court from 1945 to 2000.
    
    TORSCA                   Method to get initial starting configuration
    PRE-ITERATIONS=3         Number Iterations to Improve starting config.
    DIMMAX=2,DIMMIN=1        Maximum & Minimum Number of Dimensions
    COORDINATES=ROTATE       Rotate Coordinates so Principal Components lie along axes
    ITERATIONS=25            Maximum Number of Iterations
    REGRESSION=DESCENDING    Monotone Regression for Similarities -- NONMETRIC MDS
    DATA,LOWERHALFMATRIX,DIAGONAL=PRESENT,CUTOFF=.01 Anything below .01 is Missing Data
    U. S. SUPREME COURT AGREEMENT SCORES Title
     32  1  1                32 = # of Justices, Always set the next two numbers = 1
    (12X,101F3.0)            Format Statement For Dataset
    BURGER      100
    BLACKMUN     81100
    POWELL       86 80100
    REHNQUIS     87 72 83100
    STEVENS      71 77 74 67100
    OCONNOR      88 72 86 87 71100
    SCALIA      -99 66 85 89 65 85100
    KENNEDY     -99 70-99 88 70 86 87100         -99 is the Missing Data Code
    SOUTER      -99 72-99 78 75 81 77 84100
    THOMAS      -99 55-99 86 56 81 92 82 72100
    GINSBURG    -99 67-99 73 80 75 70 79 87 67100
    BREYER      -99-99-99 70 78 77 64 75 84 63 84100
    RUTLEDGE    -99-99-99-99-99-99-99-99-99-99-99-99100
    MURPHY      -99-99-99-99-99-99-99-99-99-99-99-99 86100
    VINSON      -99-99-99-99-99-99-99-99-99-99-99-99 63 64100
    HARLAN       81 78-99-99-99-99-99-99-99-99-99-99-99-99-99100
    BLACK        67 69-99-99-99-99-99-99-99-99-99-99 85 85 63 58100
    DOUGLAS      39 42 42 33-99-99-99-99-99-99-99-99 78 79 59 50 77100
    STEWART      77 75 80 74 75-99-99-99-99-99-99-99-99-99-99 78 67 58100
    MARSHALL     54 65 57 46 65 51 50 50 53-99-99-99-99-99-99 70 66 70 69100
    BRENNAN      53 64 56 46 65 52 51 52100-99-99-99-99-99-99 66 76 76 70 91100
    WHITE        80 76 79 77 69 77 79 80 76 74-99-99-99-99-99 74 73 56 76 59 64100
    WARREN      -99-99-99-99-99-99-99-99-99-99-99-99-99-99-99 60 81 79 71 90 91 79100
    CLARK       -99-99-99-99-99-99-99-99-99-99-99-99-99-99 91 74 67 61 77-99 77 83 77100
    FRANKFUR    -99-99-99-99-99-99-99-99-99-99-99-99 58 61 70 86 60 55 79-99 67-99 63 71100
    WHITTAKE    -99-99-99-99-99-99-99-99-99-99-99-99-99-99-99 81 57 52 82-99 66-99 62 75 80100
    BURTON      -99-99-99-99-99-99-99-99-99-99-99-99 62 58 83 77 60 56-99-99 65-99 66 81 72 80100
    REED        -99-99-99-99-99-99-99-99-99-99-99-99 65 62 84 67 60 60-99-99 69-99 71 82 67-99 82100
    FORTAS      -99-99-99-99-99-99-99-99-99-99-99-99-99-99-99 63 68 76 72 89 87 75 85 74-99-99-99-99100
    GOLDBERG    -99-99-99-99-99-99-99-99-99-99-99-99-99-99-99 59 78 80 77-99 90 78 87 71-99-99-99-99-99100
    MINTON      -99-99-99-99-99-99-99-99-99-99-99-99-99-99 87 72 62 57-99-99-99-99 75 84 68-99 82 83-99-99100
    JACKSON     -99-99-99-99-99-99-99-99-99-99-99-99 57 57 75-99 57 53-99-99-99-99 75 78 80-99 74 73-99-99 73100
    COMPUTE                 These two Lines             
    STOP                    Must Always be Included     
    
    You must run the program from a DOS Window. To run the program type:

    KYSTBIG

    The program will then prompt you for three file names: the name of the data file (it calls this the "Control Card File"); the name of an output file that you can then print out; and the name of the file for the coordinates.

    Control Card File? SUPKYST.DAT
    Printer Output File? SUPREME.PRN
    Coordinate Output File? SUPS.DAT

    The program then runs the analysis and writes the output files to disk.

    1. Produce graphs of the one and two dimensional coordinates that are in the SUPS.DAT.

    2. Interpret the one dimensional configuration. In light of what you know about the Supreme Court does in make sense to you?

  2. Download data file

    U. S. Map Driving Distance Data

    and place it in the same folder with KYST.

    The data file is reproduced below. It contains the lower half of a driving distance matrix computed between 10 U.S. cities -- Atlanta, Boise, Boston, Chicago, Cincinnati, Dallas, Denver, Los Angeles, Miami, and Washington, D.C..
    
    PRINT HISTORY, PRINT DISTANCES         This Option Prints out Some Useful Intermediate Output
    DIMMAX=3, DIMMIN=1
    TORSCA
    REGRESSION=POLYNOMIAL=1                METRIC MDS
    DATA,LOWERHALFMATRIX,DIAGONAL=PRESENT,CUTOFF=0.0
    U.S. MAP EXAMPLE
     10  1  1
    (10f5.0)
     0000
     2340 0000
     1084 2797 0000
      715 1789  976 0000
      481 2018  853  301 0000
      826 1661 1868  936  988 0000
     1519  891 2008 1017 1245  797 0000
     2252  908 3130 2189 2292 1431 1189 0000
      662 2974 1547 1386 1143 1394 2126 2885 0000
      641 2480  443  696  498 1414 1707 2754 1096 0000
    COMPUTE
    STOP
    
    1. Run this data set through KYST and get the coordinates. Plot the coordinates in two dimensions. What do you see?

    2. Change REGRESSION=POLYNOMIAL=1 to REGRESSION=ASCENDING and run it through KYST (the "ascending" tells KYST to do a Nonmetric MDS on dissimilarity data). Compare the Stress values for 1 to 3 dimensions with those obtained above and compare the two dimensional plot obtained with this option to that in part (a).

  3. Run Tables 1.1, 1.3, 1.4, and 4.1 from Borg and Groenen through KYST. Note that all four tables are similarities so that you can use the Supreme Court example as a template. For example, for Table 1.1, if you entered the data as follows:
    
     1.00
     0.52 1.00
     0.34 0.55 1.00
    etc.
    then change the format statement to:

    (7F5.2)

    Run all tables from 3 to 1 dimensions, report the Stress values, and graph the results for each table in two dimensions.