POLS 6386 MEASUREMENT THEORY
Second Assignment
Due 14 February 2001


  1. The aim of this problem is to familarize you with the Aldrich-McKelvey scaling program . Download the program and the 1968 data and "control card" files:

    MCKALNEW.F95 -- FORTRAN 95 Version of Aldrich-McKelvey scaling program. Can be compiled in gfortran

    1968 NES Data

    Urban Unrest Control Card File

    Vietnam Control Card File

    and place them in the same folder on a WINTEL machine. Read the Aldrich-McKelvey scaling program page to see how to run the program.

    Here are the first three lines of OLS68A.DAT.
    
     1681  0 10  1  1  1  1  1 63  4  4  5  7  1  2  2  3  7  1  1
     1124  0 10  1  0  0  0  1 82  1  1  4  4  1  1  1  1  4  1  5
       78  5 10  1  0  1  1  1 78  2  1  5  7  4  5  5  6  6  7  5
    
    The variables, in order, are:
    
        RESPONDENT ID      = unique 4 digit number
        PARTY ID           = 0 to 6
        RAW INCOME         = **do not use**
        FAMILY INCOME      = income quintile 1 - 5
        SEX                = 0 Man, 1 Woman
        RACE               = 0 White, 1 Black
        SOUTH              = 0 North, 1 South
        EDUCATION          = 1 High School or less, 2 Some College, 3 College
        AGE                = In Years
        URBAN UNREST SCALE = Johnson, Humphrey, Nixon, Wallace, Self-Placement
        VIETNAM SCALE      = Johnson, Humphrey, Nixon, Wallace, Self-Placement
        VOTED              = 1 Voted, 5 Did Not Vote
    
    1. Run the Aldrich-McKelvey scaling procedure using both the urban unrest and vietnam files. Save the coordinate output file. Graph the scaled stimuli coordinates against each other. Why do you think they look the way they do?

    2. Put OLS68A.DAT into Stata and define all your variables appropriately. Create the following additional variables:

      generate absparty=abs(party-3)
      generate absrespuu=abs(respuu-4)
      generate absrespvn=abs(respvn-4)

      where party is the respondent's party id, respuu is the respondent's self-placement on the urban unrest seven point scale, and respvn is the respondent's self-placement on the vietnam seven point scale. Note that these variables measure the "distance" of the respondent from the center of the respective scale.

      Test the following specification using logit and probit:

      Decision to Vote = f(income quintile, race, sex, south, education, age, absparty, absrespuu, absrespvn)

      What theory of non-voting are we testing with this specification? What should be the signs of the coefficients for the variables?

    3. Merge the files from the Aldrich-McKelvey scaling in part (a) that contain the scaled respondents into the Stata file. Note that the files will have a unique ID number for each respondent that you can use to do the file merge. Here is the first few lines of the coordinate output file for the urban unrest scale:
      
       LINE #  CASE # R POS   ALPHA    BETA    SCALED POS     RSQ
            1   1681   1.0   -2.6243    0.5249   -2.0994    0.9994    0.9997
            2   1124   1.0   -0.8831    0.3532   -0.5298    0.6790    0.8240
            3     78   4.0   -0.9588    0.2557    0.0639    0.8992    0.9483
            4    553   4.0   -1.2302    0.2895   -0.0724    0.6460    0.8037
      The second column is the respondent ID number. The columns you want to merge into the Stata File are columns two through seven (the last column is the Pearson-r).

      Create the following additional variables:

      generate absrespuu2=abs(respuu2)
      generate absrespvn2=abs(respvn2)

      where respuu2 is the respondent's scaled self-placement on the urban unrest seven point scale, and respvn2 is the respondent's scaled self-placement on the vietnam seven point scale.

      Re-run the specification in part (b) with absrespuu2 and absrespvn2 instead of absrespuu and absrespvn.

    4. Paste the Stata dataset into EVIEWS and replicate your results.