POLS 6386 MEASUREMENT THEORY
Sixth Assignment
Due 21 March 2001

1. The aim of this problem is to study the effect that being chosen as a Vice Presidential running mate has on a Senator's roll call voting record. We are going to start with Senator Richard Schweiker of Pennsylvania who was selected by Ronald Reagan to be his running mate in 1976 just prior to the Republican convention. Download the following roll call data sets:
   
   94^th Senate Roll Call Matrix
   
   95^th Senate Roll Call Matrix
   
   Combined 94^th and 95^th Senate Roll Call Matrix
   
   The combined data matrix has one record for each unique Senator who served in either or both the 94^th and 95^th Senates. In effect, we are treating the two Senates as if they were one Congress. If a member served in both Senates then his record contains both the 1311 roll calls for the 94^th plus the 1156 roll calls for the 95^th for a total of 2467 roll calls. If a member served in the 94^th but not the 95^th, then his record has the 1311 votes for the 94^th followed by 1156 zeroes (in other words, it is as if he did not vote in the 2^nd half of the "Congress"). If a member served in the 95^th but not the 94^th, then his record has 1311 zeroes for the 94^th followed by the 1156 votes in the 95^th (in other words, it is as if he did not vote in the 1^st half of the "Congress").
   
   When we use the combined matrix we are assuming that Senators do not change their ideal points between Senates!
   
   We are going to start by running these three roll call matrices through a large-matrix version of my optimal classification program -- PERFL_2006. This version has been dimensioned large enough so that we can analyze roll call matrices with up to 21000 roll call votes. Download the program and its "control card" file and place them in the same directory as the roll call votes.
   
   Large Version Maximum Classification Program -- Can be compiled with gfortran. (The R Package can be accessed from this page.)
   
   Control Card File for Classification Program
   
   The control card file for a one dimensional scaling looks like this:

   
   SEN9495.ORD
   NON-PARAMETRIC MULTIDIMENSIONAL UNFOLDING
       1 2467   10   19    2    4   20 0.025    3
   (3X,9A1,7X,4A1,2X,6A1,5X,3600I1)
   (I5,1X,19A1,2I5,50F8.3)

   1. Run PERFL_2006 in one, two, and three dimensions on all three roll call matrices. Report the one, two, and three dimensional correct classification percentages and use Excel to make a graph of the eigenvalues of the double-centered squared distance matrix computed from the agreement scores from the combined matrix.
      
      
   2. Use SPSS to make a graph of the two dimensional results of all three scalings. In the combined scaling, note that Schweiker is represented as three points! -- one constraining him to be at the same point for both Congresses, one for his voting in the 94^th only, and one for the 95^th only. Highlight Schweiker's ideal points on all of your plots.
      
      
   3. Put the coordinate output files into Excel and compute the correlations between the corresponding two and three dimensional solutions for all three matrices. Note that this will require some ingenuity on your part because not all the Senators served in both Senates and note that the rows in the combined file are slightly different from the rows in the 94^th and 95^th files. Report the correlations one dimension at a time for each dimensionality using the format below:

      
      
            One Dimension Scalings
                PERF 1d    KYST 1d         WNOM 1d
      ------------------------------------------------
      PERF 1d  1		
      KYST 1d	-0.92682	1	
      WNOM 1d	-0.95408	0.990908	1
      
      
            Two Dimensional Scalings
       	1 PERF 2d	1 KYST 2d	1 WNOM 2d
      -------------------------------------------------
      1 PERF 2d	1		
      1 KYST 2d	0.99034	1	
      1 WNOM 2d	1	0.99034 	1
      
       	2 PERF 2d	2 KYST 2d	2 WNOM 2d
      -------------------------------------------------
      2 PERF 2d	1		
      2 KYST 2d	0.592688	1	
      2 WNOM 2d	1	0.592688	1
      

2. Download HOUSYM3.FOR and its control card file below. It is a large version of the HOUSYM3 program we used in previous homeworks to compute an agreement score matrix.
   
   HOUSYM3.F95 -- FORTRAN 95 Agreement Score Program. Can easily be compiled with gfortran.
   
   symstrt3.dat -- Control Card File for HOUSYM3 Program
   
   Use HOUSYM3 to get an agreement score matrix for the 94^th, 95^th, and combined 94^th and 95^th Senates. Run the three agreement score matrices through KYST. Report the one, two, and three dimensional STRESS values for each matrix and use Excel to compute the correlations between the two and three dimensional coordinates from KYST with those from PERFL_2006.
   
   
3. In prepartion for our project to study whether or not Senator Lieberman of Connecticut changed his voting behavior because of his selection as former Vice President Gore's running mate in 2000, you need to download all the Senate roll call votes taken to date from:
   
   U.S. Congress Web Site (Thomas)
   
   Using your browser, save the roll call vote web pages as text files! Using Epsilon inset the following text at the very top of each web page:
   
   ****ROLL CALL NUMBER****
   
   followed by the vote number twice so it looks like this:

   
   ****ROLL CALL NUMBER****  12  12 Vote Number 12
   
   
   
   
   
         U.S. Senate Roll Call Votes
         107th Congress - 1st Session (2001)
         as compiled through Senate LIS by the Senate Bill Clerk 
         under the direction of the Secretary of the Senate

   This is the 12^th roll call vote. Note that the blank lines will be present when you save the web page as a text file (don't worry about the blank lines).
   
   Download the 105^th Senate roll call votes
   
   105^th Senate Roll Call Data
   
   and run PERFL_2006 in one dimension (there are 612 roll call votes). Take the rank ordering (remember that in PERF25.DAT the Senators are shown first sorted in rank order) and find the cutpoint that best classifies the Ashcroft confirmation vote (number 8, 1 February 2001). Note that you are using the 105^th Senate to classify a vote in the 107^th so that there will be a number of missing Senators. Compare your results to those I have posted on my website under "analyses of recent politics." Based upon his rank position in the 105^th, should Senator Lieberman have voted for Ashcroft? Turn in all of your calculations.